US11315658B2 - Systems and methods for deconvolution of expression data - Google Patents
Systems and methods for deconvolution of expression data Download PDFInfo
- Publication number
- US11315658B2 US11315658B2 US17/200,492 US202117200492A US11315658B2 US 11315658 B2 US11315658 B2 US 11315658B2 US 202117200492 A US202117200492 A US 202117200492A US 11315658 B2 US11315658 B2 US 11315658B2
- Authority
- US
- United States
- Prior art keywords
- cell type
- expression data
- cell
- cells
- rna
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000014509 gene expression Effects 0.000 title claims abstract description 251
- 238000000034 method Methods 0.000 title claims abstract description 82
- 108090000623 proteins and genes Proteins 0.000 claims abstract description 155
- 239000000203 mixture Substances 0.000 claims abstract description 141
- 238000012417 linear regression Methods 0.000 claims abstract description 94
- 239000012472 biological sample Substances 0.000 claims abstract description 61
- 238000012545 processing Methods 0.000 claims abstract description 12
- 210000004027 cell Anatomy 0.000 claims description 546
- 206010028980 Neoplasm Diseases 0.000 claims description 60
- 102100031381 Fc receptor-like A Human genes 0.000 claims description 38
- 101000716102 Homo sapiens T-cell surface glycoprotein CD4 Proteins 0.000 claims description 38
- 102100036011 T-cell surface glycoprotein CD4 Human genes 0.000 claims description 38
- 238000012549 training Methods 0.000 claims description 33
- 210000001744 T-lymphocyte Anatomy 0.000 claims description 32
- 201000011510 cancer Diseases 0.000 claims description 30
- 101000934346 Homo sapiens T-cell surface antigen CD2 Proteins 0.000 claims description 27
- 102100025237 T-cell surface antigen CD2 Human genes 0.000 claims description 27
- -1 WAS Proteins 0.000 claims description 27
- 101000835928 Homo sapiens Signal-regulatory protein gamma Proteins 0.000 claims description 24
- 101001050476 Homo sapiens Tyrosine-protein kinase ITK/TSK Proteins 0.000 claims description 24
- 102100025795 Signal-regulatory protein gamma Human genes 0.000 claims description 24
- 102100023345 Tyrosine-protein kinase ITK/TSK Human genes 0.000 claims description 24
- 210000001266 CD8-positive T-lymphocyte Anatomy 0.000 claims description 21
- 102100020715 Fms-related tyrosine kinase 3 ligand protein Human genes 0.000 claims description 21
- 101000932480 Homo sapiens Fms-related tyrosine kinase 3 ligand Proteins 0.000 claims description 21
- 101000934376 Homo sapiens T-cell differentiation antigen CD6 Proteins 0.000 claims description 21
- 101000934341 Homo sapiens T-cell surface glycoprotein CD5 Proteins 0.000 claims description 21
- 102100025131 T-cell differentiation antigen CD6 Human genes 0.000 claims description 21
- 102100025244 T-cell surface glycoprotein CD5 Human genes 0.000 claims description 21
- 101000634846 Homo sapiens T-cell receptor-associated transmembrane adapter 1 Proteins 0.000 claims description 20
- 101000946863 Homo sapiens T-cell surface glycoprotein CD3 delta chain Proteins 0.000 claims description 20
- 101000653540 Homo sapiens Transcription factor 7 Proteins 0.000 claims description 20
- 102100029453 T-cell receptor-associated transmembrane adapter 1 Human genes 0.000 claims description 20
- 102100030627 Transcription factor 7 Human genes 0.000 claims description 20
- 102100027205 B-cell antigen receptor complex-associated protein alpha chain Human genes 0.000 claims description 19
- 102100027203 B-cell antigen receptor complex-associated protein beta chain Human genes 0.000 claims description 19
- 102100038080 B-cell receptor CD22 Human genes 0.000 claims description 19
- 102100021568 B-cell scaffold protein with ankyrin repeats Human genes 0.000 claims description 19
- 102100024222 B-lymphocyte antigen CD19 Human genes 0.000 claims description 19
- 102100034712 C-type lectin domain family 17, member A Human genes 0.000 claims description 19
- 102100032768 Complement receptor type 2 Human genes 0.000 claims description 19
- 102100032643 Copine-5 Human genes 0.000 claims description 19
- 102100025621 Cytochrome b-245 heavy chain Human genes 0.000 claims description 19
- 102100031511 Fc receptor-like protein 2 Human genes 0.000 claims description 19
- 102100031512 Fc receptor-like protein 3 Human genes 0.000 claims description 19
- 102100031507 Fc receptor-like protein 5 Human genes 0.000 claims description 19
- 101150051800 Fcrl1 gene Proteins 0.000 claims description 19
- 101150032879 Fcrl5 gene Proteins 0.000 claims description 19
- 101150032412 Fcrla gene Proteins 0.000 claims description 19
- 101000914489 Homo sapiens B-cell antigen receptor complex-associated protein alpha chain Proteins 0.000 claims description 19
- 101000914491 Homo sapiens B-cell antigen receptor complex-associated protein beta chain Proteins 0.000 claims description 19
- 101000884305 Homo sapiens B-cell receptor CD22 Proteins 0.000 claims description 19
- 101000971155 Homo sapiens B-cell scaffold protein with ankyrin repeats Proteins 0.000 claims description 19
- 101000980825 Homo sapiens B-lymphocyte antigen CD19 Proteins 0.000 claims description 19
- 101000946279 Homo sapiens C-type lectin domain family 17, member A Proteins 0.000 claims description 19
- 101100275686 Homo sapiens CR2 gene Proteins 0.000 claims description 19
- 101000941772 Homo sapiens Copine-5 Proteins 0.000 claims description 19
- 101000846860 Homo sapiens Fc receptor-like A Proteins 0.000 claims description 19
- 101000892451 Homo sapiens Fc receptor-like B Proteins 0.000 claims description 19
- 101000846911 Homo sapiens Fc receptor-like protein 2 Proteins 0.000 claims description 19
- 101000846910 Homo sapiens Fc receptor-like protein 3 Proteins 0.000 claims description 19
- 101000777628 Homo sapiens Leukocyte antigen CD37 Proteins 0.000 claims description 19
- 101000916644 Homo sapiens Macrophage colony-stimulating factor 1 receptor Proteins 0.000 claims description 19
- 101001109501 Homo sapiens NKG2-D type II integral membrane protein Proteins 0.000 claims description 19
- 101000979455 Homo sapiens Protein Niban 3 Proteins 0.000 claims description 19
- 101000639063 Homo sapiens Protein UXT Proteins 0.000 claims description 19
- 101000648042 Homo sapiens Signal-transducing adaptor protein 1 Proteins 0.000 claims description 19
- 101000825182 Homo sapiens Transcription factor Spi-B Proteins 0.000 claims description 19
- 101000795167 Homo sapiens Tumor necrosis factor receptor superfamily member 13B Proteins 0.000 claims description 19
- 101000795169 Homo sapiens Tumor necrosis factor receptor superfamily member 13C Proteins 0.000 claims description 19
- 101000984551 Homo sapiens Tyrosine-protein kinase Blk Proteins 0.000 claims description 19
- 102100031586 Leukocyte antigen CD37 Human genes 0.000 claims description 19
- 102100028198 Macrophage colony-stimulating factor 1 receptor Human genes 0.000 claims description 19
- 101100275687 Mus musculus Cr2 gene Proteins 0.000 claims description 19
- 108010082739 NADPH Oxidase 2 Proteins 0.000 claims description 19
- 102100022680 NKG2-D type II integral membrane protein Human genes 0.000 claims description 19
- 102000036938 POU2AF1 Human genes 0.000 claims description 19
- 108060006456 POU2AF1 Proteins 0.000 claims description 19
- 102100023095 Protein Niban 3 Human genes 0.000 claims description 19
- 102100025263 Signal-transducing adaptor protein 1 Human genes 0.000 claims description 19
- 102100035891 T-cell surface glycoprotein CD3 delta chain Human genes 0.000 claims description 19
- 102100022281 Transcription factor Spi-B Human genes 0.000 claims description 19
- 102100029675 Tumor necrosis factor receptor superfamily member 13B Human genes 0.000 claims description 19
- 102100029690 Tumor necrosis factor receptor superfamily member 13C Human genes 0.000 claims description 19
- 102100027053 Tyrosine-protein kinase Blk Human genes 0.000 claims description 19
- 101000946860 Homo sapiens T-cell surface glycoprotein CD3 epsilon chain Proteins 0.000 claims description 18
- 101000738413 Homo sapiens T-cell surface glycoprotein CD3 gamma chain Proteins 0.000 claims description 18
- 101001047681 Homo sapiens Tyrosine-protein kinase Lck Proteins 0.000 claims description 18
- 101710153660 Nuclear receptor corepressor 2 Proteins 0.000 claims description 18
- 102100029452 T cell receptor alpha chain constant Human genes 0.000 claims description 18
- 102100035794 T-cell surface glycoprotein CD3 epsilon chain Human genes 0.000 claims description 18
- 102100037911 T-cell surface glycoprotein CD3 gamma chain Human genes 0.000 claims description 18
- 102100024036 Tyrosine-protein kinase Lck Human genes 0.000 claims description 18
- 210000002540 macrophage Anatomy 0.000 claims description 18
- 210000000822 natural killer cell Anatomy 0.000 claims description 18
- 102100027207 CD27 antigen Human genes 0.000 claims description 17
- 101000914511 Homo sapiens CD27 antigen Proteins 0.000 claims description 17
- 101001049181 Homo sapiens Killer cell lectin-like receptor subfamily B member 1 Proteins 0.000 claims description 17
- 101001137987 Homo sapiens Lymphocyte activation gene 3 protein Proteins 0.000 claims description 17
- 101000764535 Homo sapiens Lymphotoxin-alpha Proteins 0.000 claims description 17
- 101000714164 Homo sapiens Protein TESPA1 Proteins 0.000 claims description 17
- 101000662902 Homo sapiens T cell receptor beta constant 2 Proteins 0.000 claims description 17
- 101000914496 Homo sapiens T-cell antigen CD7 Proteins 0.000 claims description 17
- 101000671855 Homo sapiens Ubiquitin-associated and SH3 domain-containing protein A Proteins 0.000 claims description 17
- 102100023678 Killer cell lectin-like receptor subfamily B member 1 Human genes 0.000 claims description 17
- 102000017578 LAG3 Human genes 0.000 claims description 17
- 101001089108 Lotus tetragonolobus Anti-H(O) lectin Proteins 0.000 claims description 17
- 102100026238 Lymphotoxin-alpha Human genes 0.000 claims description 17
- 102100036493 Protein TESPA1 Human genes 0.000 claims description 17
- 102100037298 T cell receptor beta constant 2 Human genes 0.000 claims description 17
- 102100027208 T-cell antigen CD7 Human genes 0.000 claims description 17
- 102100040337 Ubiquitin-associated and SH3 domain-containing protein A Human genes 0.000 claims description 17
- 210000003719 b-lymphocyte Anatomy 0.000 claims description 17
- 102100031172 C-C chemokine receptor type 1 Human genes 0.000 claims description 16
- 101710149814 C-C chemokine receptor type 1 Proteins 0.000 claims description 16
- 102100040841 C-type lectin domain family 5 member A Human genes 0.000 claims description 16
- 102100025849 Complement C1q subcomponent subunit C Human genes 0.000 claims description 16
- 101000749314 Homo sapiens C-type lectin domain family 5 member A Proteins 0.000 claims description 16
- 101000933636 Homo sapiens Complement C1q subcomponent subunit C Proteins 0.000 claims description 16
- 101001055145 Homo sapiens Interleukin-2 receptor subunit beta Proteins 0.000 claims description 16
- 101000934372 Homo sapiens Macrosialin Proteins 0.000 claims description 16
- 101000956317 Homo sapiens Membrane-spanning 4-domains subfamily A member 4A Proteins 0.000 claims description 16
- 101000946889 Homo sapiens Monocyte differentiation antigen CD14 Proteins 0.000 claims description 16
- 101001097889 Homo sapiens Platelet-activating factor acetylhydrolase Proteins 0.000 claims description 16
- 101001117305 Homo sapiens Prostaglandin D2 receptor Proteins 0.000 claims description 16
- 101000868472 Homo sapiens Sialoadhesin Proteins 0.000 claims description 16
- 101001102797 Homo sapiens Transmembrane protein PVRIG Proteins 0.000 claims description 16
- 101000743488 Homo sapiens V-set and immunoglobulin domain-containing protein 4 Proteins 0.000 claims description 16
- 102100026879 Interleukin-2 receptor subunit beta Human genes 0.000 claims description 16
- 102100025136 Macrosialin Human genes 0.000 claims description 16
- 102100038556 Membrane-spanning 4-domains subfamily A member 4A Human genes 0.000 claims description 16
- 102100035877 Monocyte differentiation antigen CD14 Human genes 0.000 claims description 16
- 102100037518 Platelet-activating factor acetylhydrolase Human genes 0.000 claims description 16
- 102100024212 Prostaglandin D2 receptor Human genes 0.000 claims description 16
- 102100032855 Sialoadhesin Human genes 0.000 claims description 16
- 102100039630 Transmembrane protein PVRIG Human genes 0.000 claims description 16
- 102100038296 V-set and immunoglobulin domain-containing protein 4 Human genes 0.000 claims description 16
- 102100022005 B-lymphocyte antigen CD20 Human genes 0.000 claims description 15
- 102100031658 C-X-C chemokine receptor type 5 Human genes 0.000 claims description 15
- 102100035925 DNA methyltransferase 1-associated protein 1 Human genes 0.000 claims description 15
- 102100030442 Derlin-3 Human genes 0.000 claims description 15
- 102100028689 Glucocorticoid-induced transcript 1 protein Human genes 0.000 claims description 15
- 102100031546 HLA class II histocompatibility antigen, DO beta chain Human genes 0.000 claims description 15
- 101000897405 Homo sapiens B-lymphocyte antigen CD20 Proteins 0.000 claims description 15
- 101000922405 Homo sapiens C-X-C chemokine receptor type 5 Proteins 0.000 claims description 15
- 101000930289 Homo sapiens DNA methyltransferase 1-associated protein 1 Proteins 0.000 claims description 15
- 101000842622 Homo sapiens Derlin-3 Proteins 0.000 claims description 15
- 101000966913 Homo sapiens ELL-associated factor 2 Proteins 0.000 claims description 15
- 101001058426 Homo sapiens Glucocorticoid-induced transcript 1 protein Proteins 0.000 claims description 15
- 101000866281 Homo sapiens HLA class II histocompatibility antigen, DO beta chain Proteins 0.000 claims description 15
- 101100125778 Homo sapiens IGHM gene Proteins 0.000 claims description 15
- 101000961156 Homo sapiens Immunoglobulin heavy constant gamma 1 Proteins 0.000 claims description 15
- 101000961145 Homo sapiens Immunoglobulin heavy constant gamma 3 Proteins 0.000 claims description 15
- 101000840257 Homo sapiens Immunoglobulin kappa constant Proteins 0.000 claims description 15
- 101000840266 Homo sapiens Immunoglobulin lambda-like polypeptide 5 Proteins 0.000 claims description 15
- 101001128500 Homo sapiens Marginal zone B- and B1-cell-specific protein Proteins 0.000 claims description 15
- 101000601724 Homo sapiens Paired box protein Pax-5 Proteins 0.000 claims description 15
- 101000891028 Homo sapiens Peptidyl-prolyl cis-trans isomerase FKBP11 Proteins 0.000 claims description 15
- 101000854887 Homo sapiens Pre-B lymphocyte protein 3 Proteins 0.000 claims description 15
- 101001093096 Homo sapiens Signal peptidase complex catalytic subunit SEC11C Proteins 0.000 claims description 15
- 101000844208 Homo sapiens Thioredoxin domain-containing protein 11 Proteins 0.000 claims description 15
- 101000773122 Homo sapiens Thioredoxin domain-containing protein 5 Proteins 0.000 claims description 15
- 101000629921 Homo sapiens Translocon-associated protein subunit delta Proteins 0.000 claims description 15
- 101000801255 Homo sapiens Tumor necrosis factor receptor superfamily member 17 Proteins 0.000 claims description 15
- 102100039345 Immunoglobulin heavy constant gamma 1 Human genes 0.000 claims description 15
- 102100039348 Immunoglobulin heavy constant gamma 3 Human genes 0.000 claims description 15
- 102100039352 Immunoglobulin heavy constant mu Human genes 0.000 claims description 15
- 102100029572 Immunoglobulin kappa constant Human genes 0.000 claims description 15
- 102100029617 Immunoglobulin lambda-like polypeptide 5 Human genes 0.000 claims description 15
- 102100031826 Marginal zone B- and B1-cell-specific protein Human genes 0.000 claims description 15
- 102100037504 Paired box protein Pax-5 Human genes 0.000 claims description 15
- 102100040348 Peptidyl-prolyl cis-trans isomerase FKBP11 Human genes 0.000 claims description 15
- 102100020742 Pre-B lymphocyte protein 3 Human genes 0.000 claims description 15
- 102100036267 Signal peptidase complex catalytic subunit SEC11C Human genes 0.000 claims description 15
- 102100032030 Thioredoxin domain-containing protein 11 Human genes 0.000 claims description 15
- 102100030269 Thioredoxin domain-containing protein 5 Human genes 0.000 claims description 15
- 102100026226 Translocon-associated protein subunit delta Human genes 0.000 claims description 15
- 102100033726 Tumor necrosis factor receptor superfamily member 17 Human genes 0.000 claims description 15
- 210000001616 monocyte Anatomy 0.000 claims description 15
- 102100038077 CD226 antigen Human genes 0.000 claims description 14
- 102100035273 E3 ubiquitin-protein ligase CBL-B Human genes 0.000 claims description 14
- 101000884298 Homo sapiens CD226 antigen Proteins 0.000 claims description 14
- 101000737265 Homo sapiens E3 ubiquitin-protein ligase CBL-B Proteins 0.000 claims description 14
- 101000994669 Homo sapiens Potassium voltage-gated channel subfamily A member 3 Proteins 0.000 claims description 14
- 101000713602 Homo sapiens T-box transcription factor TBX21 Proteins 0.000 claims description 14
- 101000786318 Homo sapiens Zinc finger BED domain-containing protein 2 Proteins 0.000 claims description 14
- 102100034355 Potassium voltage-gated channel subfamily A member 3 Human genes 0.000 claims description 14
- 102100036840 T-box transcription factor TBX21 Human genes 0.000 claims description 14
- 102100025797 Zinc finger BED domain-containing protein 2 Human genes 0.000 claims description 14
- 210000002950 fibroblast Anatomy 0.000 claims description 14
- 230000003211 malignant effect Effects 0.000 claims description 14
- 210000000440 neutrophil Anatomy 0.000 claims description 14
- 102100024358 Arf-GAP with dual PH domain-containing protein 2 Human genes 0.000 claims description 13
- 102100032367 C-C motif chemokine 5 Human genes 0.000 claims description 13
- 102100032366 C-C motif chemokine 7 Human genes 0.000 claims description 13
- 102100021703 C3a anaphylatoxin chemotactic receptor Human genes 0.000 claims description 13
- 102100032957 C5a anaphylatoxin chemotactic receptor 1 Human genes 0.000 claims description 13
- 102100024263 CD160 antigen Human genes 0.000 claims description 13
- 108010009992 CD163 antigen Proteins 0.000 claims description 13
- 102100031011 Chemerin-like receptor 1 Human genes 0.000 claims description 13
- 102100035298 Cytokine SCM-1 beta Human genes 0.000 claims description 13
- 102100022087 Granzyme M Human genes 0.000 claims description 13
- 101000832784 Homo sapiens Arf-GAP with dual PH domain-containing protein 2 Proteins 0.000 claims description 13
- 101000797762 Homo sapiens C-C motif chemokine 5 Proteins 0.000 claims description 13
- 101000797758 Homo sapiens C-C motif chemokine 7 Proteins 0.000 claims description 13
- 101000896583 Homo sapiens C3a anaphylatoxin chemotactic receptor Proteins 0.000 claims description 13
- 101000867983 Homo sapiens C5a anaphylatoxin chemotactic receptor 1 Proteins 0.000 claims description 13
- 101000761938 Homo sapiens CD160 antigen Proteins 0.000 claims description 13
- 101000919756 Homo sapiens Chemerin-like receptor 1 Proteins 0.000 claims description 13
- 101000804771 Homo sapiens Cytokine SCM-1 beta Proteins 0.000 claims description 13
- 101000900697 Homo sapiens Granzyme M Proteins 0.000 claims description 13
- 101000971538 Homo sapiens Killer cell lectin-like receptor subfamily F member 1 Proteins 0.000 claims description 13
- 101001121408 Homo sapiens L-amino-acid oxidase Proteins 0.000 claims description 13
- 101001134216 Homo sapiens Macrophage scavenger receptor types I and II Proteins 0.000 claims description 13
- 101000934338 Homo sapiens Myeloid cell surface antigen CD33 Proteins 0.000 claims description 13
- 101001059802 Homo sapiens N-formyl peptide receptor 3 Proteins 0.000 claims description 13
- 101001109503 Homo sapiens NKG2-C type II integral membrane protein Proteins 0.000 claims description 13
- 101000884270 Homo sapiens Natural killer cell receptor 2B4 Proteins 0.000 claims description 13
- 101000987581 Homo sapiens Perforin-1 Proteins 0.000 claims description 13
- 101000653759 Homo sapiens Sphingosine 1-phosphate receptor 5 Proteins 0.000 claims description 13
- 101000890836 Homo sapiens TRAF3-interacting JNK-activating modulator Proteins 0.000 claims description 13
- 101000764622 Homo sapiens Transmembrane and immunoglobulin domain-containing protein 2 Proteins 0.000 claims description 13
- 101000795117 Homo sapiens Triggering receptor expressed on myeloid cells 2 Proteins 0.000 claims description 13
- 108090000174 Interleukin-10 Proteins 0.000 claims description 13
- 102000003814 Interleukin-10 Human genes 0.000 claims description 13
- 102100021458 Killer cell lectin-like receptor subfamily F member 1 Human genes 0.000 claims description 13
- 102100026388 L-amino-acid oxidase Human genes 0.000 claims description 13
- 102100025354 Macrophage mannose receptor 1 Human genes 0.000 claims description 13
- 102100034184 Macrophage scavenger receptor types I and II Human genes 0.000 claims description 13
- 108010031099 Mannose Receptor Proteins 0.000 claims description 13
- 102100025243 Myeloid cell surface antigen CD33 Human genes 0.000 claims description 13
- 102100028130 N-formyl peptide receptor 3 Human genes 0.000 claims description 13
- 102100022683 NKG2-C type II integral membrane protein Human genes 0.000 claims description 13
- 102100038082 Natural killer cell receptor 2B4 Human genes 0.000 claims description 13
- 102100028467 Perforin-1 Human genes 0.000 claims description 13
- 108091000520 Protein-Arginine Deiminase Type 4 Proteins 0.000 claims description 13
- 102100035731 Protein-arginine deiminase type-4 Human genes 0.000 claims description 13
- 108010059278 Pyrin Proteins 0.000 claims description 13
- 102100039233 Pyrin Human genes 0.000 claims description 13
- 102100025831 Scavenger receptor cysteine-rich type 1 protein M130 Human genes 0.000 claims description 13
- 102100029802 Sphingosine 1-phosphate receptor 5 Human genes 0.000 claims description 13
- 102100040128 TRAF3-interacting JNK-activating modulator Human genes 0.000 claims description 13
- 102100026224 Transmembrane and immunoglobulin domain-containing protein 2 Human genes 0.000 claims description 13
- 102100029678 Triggering receptor expressed on myeloid cells 2 Human genes 0.000 claims description 13
- 210000004698 lymphocyte Anatomy 0.000 claims description 12
- 238000003860 storage Methods 0.000 claims description 12
- 102100036008 CD48 antigen Human genes 0.000 claims description 11
- 102100022789 Calcium/calmodulin-dependent protein kinase type IV Human genes 0.000 claims description 11
- 101000716130 Homo sapiens CD48 antigen Proteins 0.000 claims description 11
- 101000974816 Homo sapiens Calcium/calmodulin-dependent protein kinase type IV Proteins 0.000 claims description 11
- 101001047640 Homo sapiens Linker for activation of T-cells family member 1 Proteins 0.000 claims description 11
- 101000702132 Homo sapiens Protein spinster homolog 1 Proteins 0.000 claims description 11
- 101000689224 Homo sapiens Src-like-adapter 2 Proteins 0.000 claims description 11
- 101000798076 Homo sapiens T cell receptor delta constant Proteins 0.000 claims description 11
- 101000679306 Homo sapiens T cell receptor gamma constant 1 Proteins 0.000 claims description 11
- 101000679307 Homo sapiens T cell receptor gamma constant 2 Proteins 0.000 claims description 11
- 101000946843 Homo sapiens T-cell surface glycoprotein CD8 alpha chain Proteins 0.000 claims description 11
- 102100024032 Linker for activation of T-cells family member 1 Human genes 0.000 claims description 11
- 102100024510 Src-like-adapter 2 Human genes 0.000 claims description 11
- 102100032272 T cell receptor delta constant Human genes 0.000 claims description 11
- 102100022590 T cell receptor gamma constant 1 Human genes 0.000 claims description 11
- 102100022571 T cell receptor gamma constant 2 Human genes 0.000 claims description 11
- 102100034922 T-cell surface glycoprotein CD8 alpha chain Human genes 0.000 claims description 11
- 101001038499 Yarrowia lipolytica (strain CLIB 122 / E 150) Lysine acetyltransferase Proteins 0.000 claims description 11
- FDFPSNISSMYYDS-UHFFFAOYSA-N 2-ethyl-N,2-dimethylheptanamide Chemical compound CCCCCC(C)(CC)C(=O)NC FDFPSNISSMYYDS-UHFFFAOYSA-N 0.000 claims description 10
- 108091005560 ADGRG3 Proteins 0.000 claims description 10
- 102100022781 ATP-sensitive inward rectifier potassium channel 15 Human genes 0.000 claims description 10
- 102100039140 Acyloxyacyl hydrolase Human genes 0.000 claims description 10
- 102100025976 Adenosine deaminase 2 Human genes 0.000 claims description 10
- 102100036006 Adenosine receptor A3 Human genes 0.000 claims description 10
- 102100026425 Adhesion G protein-coupled receptor E3 Human genes 0.000 claims description 10
- 102100040037 Adhesion G protein-coupled receptor G3 Human genes 0.000 claims description 10
- 102100036013 Antigen-presenting glycoprotein CD1d Human genes 0.000 claims description 10
- 102100031151 C-C chemokine receptor type 2 Human genes 0.000 claims description 10
- 101710149815 C-C chemokine receptor type 2 Proteins 0.000 claims description 10
- 102100024167 C-C chemokine receptor type 3 Human genes 0.000 claims description 10
- 101710149862 C-C chemokine receptor type 3 Proteins 0.000 claims description 10
- 102100036166 C-X-C chemokine receptor type 1 Human genes 0.000 claims description 10
- 102100028989 C-X-C chemokine receptor type 2 Human genes 0.000 claims description 10
- 102100040840 C-type lectin domain family 7 member A Human genes 0.000 claims description 10
- 102100025238 CD302 antigen Human genes 0.000 claims description 10
- 102100029380 CMRF35-like molecule 2 Human genes 0.000 claims description 10
- 102100029382 CMRF35-like molecule 6 Human genes 0.000 claims description 10
- 102100029400 CMRF35-like molecule 7 Human genes 0.000 claims description 10
- 102100025470 Carcinoembryonic antigen-related cell adhesion molecule 8 Human genes 0.000 claims description 10
- 102100035654 Cathepsin S Human genes 0.000 claims description 10
- 102100026658 Cathepsin W Human genes 0.000 claims description 10
- 102100026093 Claudin domain-containing protein 2 Human genes 0.000 claims description 10
- 102100024508 Ficolin-1 Human genes 0.000 claims description 10
- 102100040133 Free fatty acid receptor 2 Human genes 0.000 claims description 10
- 102100040139 Glycosyltransferase 1 domain-containing protein 1 Human genes 0.000 claims description 10
- 102100039622 Granulocyte colony-stimulating factor receptor Human genes 0.000 claims description 10
- 101001047184 Homo sapiens ATP-sensitive inward rectifier potassium channel 15 Proteins 0.000 claims description 10
- 101000889541 Homo sapiens Acyloxyacyl hydrolase Proteins 0.000 claims description 10
- 101000720051 Homo sapiens Adenosine deaminase 2 Proteins 0.000 claims description 10
- 101000783645 Homo sapiens Adenosine receptor A3 Proteins 0.000 claims description 10
- 101000718235 Homo sapiens Adhesion G protein-coupled receptor E3 Proteins 0.000 claims description 10
- 101000716121 Homo sapiens Antigen-presenting glycoprotein CD1d Proteins 0.000 claims description 10
- 101000947174 Homo sapiens C-X-C chemokine receptor type 1 Proteins 0.000 claims description 10
- 101000749325 Homo sapiens C-type lectin domain family 7 member A Proteins 0.000 claims description 10
- 101000934351 Homo sapiens CD302 antigen Proteins 0.000 claims description 10
- 101000990046 Homo sapiens CMRF35-like molecule 2 Proteins 0.000 claims description 10
- 101000990034 Homo sapiens CMRF35-like molecule 6 Proteins 0.000 claims description 10
- 101000990007 Homo sapiens CMRF35-like molecule 7 Proteins 0.000 claims description 10
- 101000914320 Homo sapiens Carcinoembryonic antigen-related cell adhesion molecule 8 Proteins 0.000 claims description 10
- 101000947086 Homo sapiens Cathepsin S Proteins 0.000 claims description 10
- 101000910988 Homo sapiens Cathepsin W Proteins 0.000 claims description 10
- 101000912641 Homo sapiens Claudin domain-containing protein 2 Proteins 0.000 claims description 10
- 101001052785 Homo sapiens Ficolin-1 Proteins 0.000 claims description 10
- 101000890668 Homo sapiens Free fatty acid receptor 2 Proteins 0.000 claims description 10
- 101001037042 Homo sapiens Glycosyltransferase 1 domain-containing protein 1 Proteins 0.000 claims description 10
- 101000746364 Homo sapiens Granulocyte colony-stimulating factor receptor Proteins 0.000 claims description 10
- 101001011442 Homo sapiens Interferon regulatory factor 5 Proteins 0.000 claims description 10
- 101001034842 Homo sapiens Interferon-induced transmembrane protein 2 Proteins 0.000 claims description 10
- 101001055144 Homo sapiens Interleukin-2 receptor subunit alpha Proteins 0.000 claims description 10
- 101000945371 Homo sapiens Killer cell immunoglobulin-like receptor 2DL2 Proteins 0.000 claims description 10
- 101000945333 Homo sapiens Killer cell immunoglobulin-like receptor 2DL3 Proteins 0.000 claims description 10
- 101000945331 Homo sapiens Killer cell immunoglobulin-like receptor 2DL4 Proteins 0.000 claims description 10
- 101000945339 Homo sapiens Killer cell immunoglobulin-like receptor 2DS2 Proteins 0.000 claims description 10
- 101000945351 Homo sapiens Killer cell immunoglobulin-like receptor 3DL1 Proteins 0.000 claims description 10
- 101000945490 Homo sapiens Killer cell immunoglobulin-like receptor 3DL2 Proteins 0.000 claims description 10
- 101001005166 Homo sapiens Lens fiber membrane intrinsic protein Proteins 0.000 claims description 10
- 101000917826 Homo sapiens Low affinity immunoglobulin gamma Fc region receptor II-a Proteins 0.000 claims description 10
- 101000917839 Homo sapiens Low affinity immunoglobulin gamma Fc region receptor III-B Proteins 0.000 claims description 10
- 101000956320 Homo sapiens Membrane-spanning 4-domains subfamily A member 6A Proteins 0.000 claims description 10
- 101000577891 Homo sapiens Myeloid cell nuclear differentiation antigen Proteins 0.000 claims description 10
- 101000818546 Homo sapiens N-formyl peptide receptor 2 Proteins 0.000 claims description 10
- 101001109470 Homo sapiens NKG2-E type II integral membrane protein Proteins 0.000 claims description 10
- 101000589301 Homo sapiens Natural cytotoxicity triggering receptor 1 Proteins 0.000 claims description 10
- 101000589307 Homo sapiens Natural cytotoxicity triggering receptor 3 Proteins 0.000 claims description 10
- 101000971513 Homo sapiens Natural killer cells antigen CD94 Proteins 0.000 claims description 10
- 101000581981 Homo sapiens Neural cell adhesion molecule 1 Proteins 0.000 claims description 10
- 101001024391 Homo sapiens Neuromedin-U receptor 1 Proteins 0.000 claims description 10
- 101001120082 Homo sapiens P2Y purinoceptor 13 Proteins 0.000 claims description 10
- 101000986810 Homo sapiens P2Y purinoceptor 8 Proteins 0.000 claims description 10
- 101000854774 Homo sapiens Pantetheine hydrolase VNN2 Proteins 0.000 claims description 10
- 101000731015 Homo sapiens Peptidoglycan recognition protein 1 Proteins 0.000 claims description 10
- 101001071167 Homo sapiens Phosphoethanolamine/phosphocholine phosphatase Proteins 0.000 claims description 10
- 101001057168 Homo sapiens Protein EVI2B Proteins 0.000 claims description 10
- 101000979599 Homo sapiens Protein NKG7 Proteins 0.000 claims description 10
- 101001082184 Homo sapiens Pyrin and HIN domain-containing protein 1 Proteins 0.000 claims description 10
- 101000870953 Homo sapiens RAS guanyl-releasing protein 4 Proteins 0.000 claims description 10
- 101000584702 Homo sapiens Ras-related protein Rab-7b Proteins 0.000 claims description 10
- 101000707218 Homo sapiens SH2 domain-containing protein 1B Proteins 0.000 claims description 10
- 101000863900 Homo sapiens Sialic acid-binding Ig-like lectin 5 Proteins 0.000 claims description 10
- 101000688930 Homo sapiens Signaling threshold-regulating transmembrane adapter 1 Proteins 0.000 claims description 10
- 101000740162 Homo sapiens Sodium- and chloride-dependent transporter XTRP3 Proteins 0.000 claims description 10
- 101000822549 Homo sapiens Sterile alpha motif domain-containing protein 3 Proteins 0.000 claims description 10
- 101000738335 Homo sapiens T-cell surface glycoprotein CD3 zeta chain Proteins 0.000 claims description 10
- 101000914514 Homo sapiens T-cell-specific surface glycoprotein CD28 Proteins 0.000 claims description 10
- 101000610602 Homo sapiens Tumor necrosis factor receptor superfamily member 10C Proteins 0.000 claims description 10
- 101000818522 Homo sapiens fMet-Leu-Phe receptor Proteins 0.000 claims description 10
- 102100030131 Interferon regulatory factor 5 Human genes 0.000 claims description 10
- 102100040020 Interferon-induced transmembrane protein 2 Human genes 0.000 claims description 10
- 102100026878 Interleukin-2 receptor subunit alpha Human genes 0.000 claims description 10
- 108010018951 Interleukin-8B Receptors Proteins 0.000 claims description 10
- 102100033599 Killer cell immunoglobulin-like receptor 2DL2 Human genes 0.000 claims description 10
- 102100033634 Killer cell immunoglobulin-like receptor 2DL3 Human genes 0.000 claims description 10
- 102100033633 Killer cell immunoglobulin-like receptor 2DL4 Human genes 0.000 claims description 10
- 102100033630 Killer cell immunoglobulin-like receptor 2DS2 Human genes 0.000 claims description 10
- 102100033627 Killer cell immunoglobulin-like receptor 3DL1 Human genes 0.000 claims description 10
- 102100034840 Killer cell immunoglobulin-like receptor 3DL2 Human genes 0.000 claims description 10
- 102100026038 Lens fiber membrane intrinsic protein Human genes 0.000 claims description 10
- 102100029204 Low affinity immunoglobulin gamma Fc region receptor II-a Human genes 0.000 claims description 10
- 102100029185 Low affinity immunoglobulin gamma Fc region receptor III-B Human genes 0.000 claims description 10
- 102100024131 Matrix metalloproteinase-25 Human genes 0.000 claims description 10
- 102100038555 Membrane-spanning 4-domains subfamily A member 6A Human genes 0.000 claims description 10
- 101100372838 Mus musculus Vnn3 gene Proteins 0.000 claims description 10
- 102100027994 Myeloid cell nuclear differentiation antigen Human genes 0.000 claims description 10
- 102100021126 N-formyl peptide receptor 2 Human genes 0.000 claims description 10
- 102100022701 NKG2-E type II integral membrane protein Human genes 0.000 claims description 10
- 102100032870 Natural cytotoxicity triggering receptor 1 Human genes 0.000 claims description 10
- 102100032852 Natural cytotoxicity triggering receptor 3 Human genes 0.000 claims description 10
- 102100021462 Natural killer cells antigen CD94 Human genes 0.000 claims description 10
- 102100027347 Neural cell adhesion molecule 1 Human genes 0.000 claims description 10
- 102100035314 Neuromedin-U receptor 1 Human genes 0.000 claims description 10
- 102100026168 P2Y purinoceptor 13 Human genes 0.000 claims description 10
- 102100028069 P2Y purinoceptor 8 Human genes 0.000 claims description 10
- 102100020748 Pantetheine hydrolase VNN2 Human genes 0.000 claims description 10
- 102100032393 Peptidoglycan recognition protein 1 Human genes 0.000 claims description 10
- 102100036844 Phosphoethanolamine/phosphocholine phosphatase Human genes 0.000 claims description 10
- 102100027249 Protein EVI2B Human genes 0.000 claims description 10
- 102100023370 Protein NKG7 Human genes 0.000 claims description 10
- 108091000521 Protein-Arginine Deiminase Type 2 Proteins 0.000 claims description 10
- 102100035735 Protein-arginine deiminase type-2 Human genes 0.000 claims description 10
- 102100027365 Pyrin and HIN domain-containing protein 1 Human genes 0.000 claims description 10
- 102100033445 RAS guanyl-releasing protein 4 Human genes 0.000 claims description 10
- 102100030008 Ras-related protein Rab-7b Human genes 0.000 claims description 10
- 102100031778 SH2 domain-containing protein 1B Human genes 0.000 claims description 10
- 102100029957 Sialic acid-binding Ig-like lectin 5 Human genes 0.000 claims description 10
- 102100024453 Signaling threshold-regulating transmembrane adapter 1 Human genes 0.000 claims description 10
- 102100022468 Sterile alpha motif domain-containing protein 3 Human genes 0.000 claims description 10
- 102100037906 T-cell surface glycoprotein CD3 zeta chain Human genes 0.000 claims description 10
- 102100027213 T-cell-specific surface glycoprotein CD28 Human genes 0.000 claims description 10
- 102100040115 Tumor necrosis factor receptor superfamily member 10C Human genes 0.000 claims description 10
- 102100021145 fMet-Leu-Phe receptor Human genes 0.000 claims description 10
- 108090000440 matrix metalloproteinase 25 Proteins 0.000 claims description 10
- 102100026234 Cytokine receptor common subunit gamma Human genes 0.000 claims description 9
- 101001055227 Homo sapiens Cytokine receptor common subunit gamma Proteins 0.000 claims description 9
- 210000002865 immune cell Anatomy 0.000 claims description 9
- 210000002889 endothelial cell Anatomy 0.000 claims description 8
- 102100040121 Allograft inflammatory factor 1 Human genes 0.000 claims description 7
- 102100031326 Ankyrin repeat domain-containing protein 55 Human genes 0.000 claims description 7
- 102100037674 Bis(5'-adenosyl)-triphosphatase Human genes 0.000 claims description 7
- 101710149863 C-C chemokine receptor type 4 Proteins 0.000 claims description 7
- 102100026197 C-type lectin domain family 2 member D Human genes 0.000 claims description 7
- 102100032976 CCR4-NOT transcription complex subunit 6 Human genes 0.000 claims description 7
- 102100022436 CMRF35-like molecule 8 Human genes 0.000 claims description 7
- 102100029876 Capping protein, Arp2/3 and myosin-I linker protein 2 Human genes 0.000 claims description 7
- 102100026247 Carabin Human genes 0.000 claims description 7
- 102100038418 Cytoplasmic FMR1-interacting protein 2 Human genes 0.000 claims description 7
- 102100040264 DNA dC->dU-editing enzyme APOBEC-3D Human genes 0.000 claims description 7
- 102100025137 Early activation antigen CD69 Human genes 0.000 claims description 7
- 101000823089 Equus caballus Alpha-1-antiproteinase 1 Proteins 0.000 claims description 7
- 102100037815 Fas apoptotic inhibitory molecule 3 Human genes 0.000 claims description 7
- 102100040225 Gamma-interferon-inducible lysosomal thiol reductase Human genes 0.000 claims description 7
- 102100028113 Granulocyte-macrophage colony-stimulating factor receptor subunit alpha Human genes 0.000 claims description 7
- 102100022132 High affinity immunoglobulin epsilon receptor subunit gamma Human genes 0.000 claims description 7
- 102100026122 High affinity immunoglobulin gamma Fc receptor I Human genes 0.000 claims description 7
- 102100026119 High affinity immunoglobulin gamma Fc receptor IB Human genes 0.000 claims description 7
- 101000890626 Homo sapiens Allograft inflammatory factor 1 Proteins 0.000 claims description 7
- 101000796113 Homo sapiens Ankyrin repeat domain-containing protein 55 Proteins 0.000 claims description 7
- 101000912615 Homo sapiens C-type lectin domain family 2 member D Proteins 0.000 claims description 7
- 101000901669 Homo sapiens CMRF35-like molecule 8 Proteins 0.000 claims description 7
- 101000793825 Homo sapiens Capping protein, Arp2/3 and myosin-I linker protein 2 Proteins 0.000 claims description 7
- 101000835644 Homo sapiens Carabin Proteins 0.000 claims description 7
- 101000956870 Homo sapiens Cytoplasmic FMR1-interacting protein 2 Proteins 0.000 claims description 7
- 101000964382 Homo sapiens DNA dC->dU-editing enzyme APOBEC-3D Proteins 0.000 claims description 7
- 101000934374 Homo sapiens Early activation antigen CD69 Proteins 0.000 claims description 7
- 101000878510 Homo sapiens Fas apoptotic inhibitory molecule 3 Proteins 0.000 claims description 7
- 101001037132 Homo sapiens Gamma-interferon-inducible lysosomal thiol reductase Proteins 0.000 claims description 7
- 101000916625 Homo sapiens Granulocyte-macrophage colony-stimulating factor receptor subunit alpha Proteins 0.000 claims description 7
- 101000824104 Homo sapiens High affinity immunoglobulin epsilon receptor subunit gamma Proteins 0.000 claims description 7
- 101000913074 Homo sapiens High affinity immunoglobulin gamma Fc receptor I Proteins 0.000 claims description 7
- 101000913077 Homo sapiens High affinity immunoglobulin gamma Fc receptor IB Proteins 0.000 claims description 7
- 101001046668 Homo sapiens Integrin alpha-X Proteins 0.000 claims description 7
- 101001015037 Homo sapiens Integrin beta-7 Proteins 0.000 claims description 7
- 101001011441 Homo sapiens Interferon regulatory factor 4 Proteins 0.000 claims description 7
- 101001039157 Homo sapiens Leucine-rich repeat-containing protein 25 Proteins 0.000 claims description 7
- 101001047659 Homo sapiens Lymphocyte transmembrane adapter 1 Proteins 0.000 claims description 7
- 101001059991 Homo sapiens Mitogen-activated protein kinase kinase kinase kinase 1 Proteins 0.000 claims description 7
- 101000979575 Homo sapiens NLR family CARD domain-containing protein 3 Proteins 0.000 claims description 7
- 101001112224 Homo sapiens Neutrophil cytosol factor 2 Proteins 0.000 claims description 7
- 101000836112 Homo sapiens Nuclear body protein SP140 Proteins 0.000 claims description 7
- 101001129851 Homo sapiens Paired immunoglobulin-like type 2 receptor alpha Proteins 0.000 claims description 7
- 101001040717 Homo sapiens Probable G-protein coupled receptor 174 Proteins 0.000 claims description 7
- 101001122742 Homo sapiens Protein phosphatase 1 regulatory inhibitor subunit 16B Proteins 0.000 claims description 7
- 101001135804 Homo sapiens Protein tyrosine phosphatase receptor type C-associated protein Proteins 0.000 claims description 7
- 101001120091 Homo sapiens Putative P2Y purinoceptor 10 Proteins 0.000 claims description 7
- 101001069891 Homo sapiens RAS guanyl-releasing protein 1 Proteins 0.000 claims description 7
- 101000905936 Homo sapiens RAS guanyl-releasing protein 2 Proteins 0.000 claims description 7
- 101000666634 Homo sapiens Rho-related GTP-binding protein RhoH Proteins 0.000 claims description 7
- 101000633786 Homo sapiens SLAM family member 6 Proteins 0.000 claims description 7
- 101000863883 Homo sapiens Sialic acid-binding Ig-like lectin 9 Proteins 0.000 claims description 7
- 101000709268 Homo sapiens Signal-regulatory protein beta-2 Proteins 0.000 claims description 7
- 101001018021 Homo sapiens T-lymphocyte surface antigen Ly-9 Proteins 0.000 claims description 7
- 101000809875 Homo sapiens TYRO protein tyrosine kinase-binding protein Proteins 0.000 claims description 7
- 101000831567 Homo sapiens Toll-like receptor 2 Proteins 0.000 claims description 7
- 101000651211 Homo sapiens Transcription factor PU.1 Proteins 0.000 claims description 7
- 101001009087 Homo sapiens Tyrosine-protein kinase HCK Proteins 0.000 claims description 7
- 101000599042 Homo sapiens Zinc finger protein Aiolos Proteins 0.000 claims description 7
- 102100022297 Integrin alpha-X Human genes 0.000 claims description 7
- 102100033016 Integrin beta-7 Human genes 0.000 claims description 7
- 102100030126 Interferon regulatory factor 4 Human genes 0.000 claims description 7
- 101800003050 Interleukin-16 Proteins 0.000 claims description 7
- 102100040695 Leucine-rich repeat-containing protein 25 Human genes 0.000 claims description 7
- 102100024034 Lymphocyte transmembrane adapter 1 Human genes 0.000 claims description 7
- 102100028199 Mitogen-activated protein kinase kinase kinase kinase 1 Human genes 0.000 claims description 7
- 108010018525 NFATC Transcription Factors Proteins 0.000 claims description 7
- 102100023382 NLR family CARD domain-containing protein 3 Human genes 0.000 claims description 7
- 102100023618 Neutrophil cytosol factor 2 Human genes 0.000 claims description 7
- 102100025638 Nuclear body protein SP140 Human genes 0.000 claims description 7
- 102100031651 Paired immunoglobulin-like type 2 receptor alpha Human genes 0.000 claims description 7
- 102100021199 Probable G-protein coupled receptor 174 Human genes 0.000 claims description 7
- 102100028740 Protein phosphatase 1 regulatory inhibitor subunit 16B Human genes 0.000 claims description 7
- 102100036937 Protein tyrosine phosphatase receptor type C-associated protein Human genes 0.000 claims description 7
- 102100026173 Putative P2Y purinoceptor 10 Human genes 0.000 claims description 7
- 102100034220 RAS guanyl-releasing protein 1 Human genes 0.000 claims description 7
- 102100023488 RAS guanyl-releasing protein 2 Human genes 0.000 claims description 7
- 102100038338 Rho-related GTP-binding protein RhoH Human genes 0.000 claims description 7
- 102100029197 SLAM family member 6 Human genes 0.000 claims description 7
- 102100029965 Sialic acid-binding Ig-like lectin 9 Human genes 0.000 claims description 7
- 102100033447 T-lymphocyte surface antigen Ly-9 Human genes 0.000 claims description 7
- 102100038717 TYRO protein tyrosine kinase-binding protein Human genes 0.000 claims description 7
- 102100024333 Toll-like receptor 2 Human genes 0.000 claims description 7
- 102100027654 Transcription factor PU.1 Human genes 0.000 claims description 7
- 102100027389 Tyrosine-protein kinase HCK Human genes 0.000 claims description 7
- 102100037798 Zinc finger protein Aiolos Human genes 0.000 claims description 7
- 108010005713 bis(5'-adenosyl)triphosphatase Proteins 0.000 claims description 7
- 101000611936 Homo sapiens Programmed cell death protein 1 Proteins 0.000 claims description 6
- 101000738771 Homo sapiens Receptor-type tyrosine-protein phosphatase C Proteins 0.000 claims description 6
- 101000851515 Homo sapiens Transmembrane channel-like protein 8 Proteins 0.000 claims description 6
- 102100040678 Programmed cell death protein 1 Human genes 0.000 claims description 6
- 102100037422 Receptor-type tyrosine-protein phosphatase C Human genes 0.000 claims description 6
- 102100036770 Transmembrane channel-like protein 8 Human genes 0.000 claims description 6
- 210000003289 regulatory T cell Anatomy 0.000 claims description 6
- 102100027385 Hematopoietic lineage cell-specific protein Human genes 0.000 claims description 5
- 101001009091 Homo sapiens Hematopoietic lineage cell-specific protein Proteins 0.000 claims description 5
- 210000003038 endothelium Anatomy 0.000 claims description 5
- 102100030389 1-phosphatidylinositol 4,5-bisphosphate phosphodiesterase beta-2 Human genes 0.000 claims description 4
- 101150075418 ARHGAP15 gene Proteins 0.000 claims description 4
- 102100025218 B-cell differentiation antigen CD72 Human genes 0.000 claims description 4
- 102100022976 B-cell lymphoma/leukemia 11A Human genes 0.000 claims description 4
- 102100033210 CUGBP Elav-like family member 2 Human genes 0.000 claims description 4
- 101001110283 Canis lupus familiaris Ras-related C3 botulinum toxin substrate 1 Proteins 0.000 claims description 4
- 102100032678 CapZ-interacting protein Human genes 0.000 claims description 4
- 102100034951 Coiled-coil domain-containing protein 69 Human genes 0.000 claims description 4
- 102100028233 Coronin-1A Human genes 0.000 claims description 4
- 102100025843 Cytohesin-4 Human genes 0.000 claims description 4
- 102100028183 Cytohesin-interacting protein Human genes 0.000 claims description 4
- 102100033507 DENN domain-containing protein 1C Human genes 0.000 claims description 4
- 102100031597 Dedicator of cytokinesis protein 2 Human genes 0.000 claims description 4
- 102100036727 Deformed epidermal autoregulatory factor 1 homolog Human genes 0.000 claims description 4
- 102100022820 Disintegrin and metalloproteinase domain-containing protein 28 Human genes 0.000 claims description 4
- 102100027118 Engulfment and cell motility protein 1 Human genes 0.000 claims description 4
- 101000914063 Eucalyptus globulus Leafy/floricaula homolog FL1 Proteins 0.000 claims description 4
- 102100040612 Fermitin family homolog 3 Human genes 0.000 claims description 4
- 102100040680 Formin-binding protein 1 Human genes 0.000 claims description 4
- 102100028930 Formin-like protein 1 Human genes 0.000 claims description 4
- 102100040903 Gamma-parvin Human genes 0.000 claims description 4
- 101000583066 Homo sapiens 1-phosphatidylinositol 4,5-bisphosphate phosphodiesterase beta-2 Proteins 0.000 claims description 4
- 101000934359 Homo sapiens B-cell differentiation antigen CD72 Proteins 0.000 claims description 4
- 101000903703 Homo sapiens B-cell lymphoma/leukemia 11A Proteins 0.000 claims description 4
- 101000944442 Homo sapiens CUGBP Elav-like family member 2 Proteins 0.000 claims description 4
- 101000941906 Homo sapiens CapZ-interacting protein Proteins 0.000 claims description 4
- 101000946601 Homo sapiens Coiled-coil domain-containing protein 69 Proteins 0.000 claims description 4
- 101000860852 Homo sapiens Coronin-1A Proteins 0.000 claims description 4
- 101000855828 Homo sapiens Cytohesin-4 Proteins 0.000 claims description 4
- 101000916686 Homo sapiens Cytohesin-interacting protein Proteins 0.000 claims description 4
- 101000870874 Homo sapiens DENN domain-containing protein 1C Proteins 0.000 claims description 4
- 101000866237 Homo sapiens Dedicator of cytokinesis protein 2 Proteins 0.000 claims description 4
- 101000929421 Homo sapiens Deformed epidermal autoregulatory factor 1 homolog Proteins 0.000 claims description 4
- 101000756756 Homo sapiens Disintegrin and metalloproteinase domain-containing protein 28 Proteins 0.000 claims description 4
- 101000877395 Homo sapiens ETS-related transcription factor Elf-1 Proteins 0.000 claims description 4
- 101001057862 Homo sapiens Engulfment and cell motility protein 1 Proteins 0.000 claims description 4
- 101000749644 Homo sapiens Fermitin family homolog 3 Proteins 0.000 claims description 4
- 101000892722 Homo sapiens Formin-binding protein 1 Proteins 0.000 claims description 4
- 101001059386 Homo sapiens Formin-like protein 1 Proteins 0.000 claims description 4
- 101000613555 Homo sapiens Gamma-parvin Proteins 0.000 claims description 4
- 101001076297 Homo sapiens IGF-like family receptor 1 Proteins 0.000 claims description 4
- 101001046683 Homo sapiens Integrin alpha-L Proteins 0.000 claims description 4
- 101000935040 Homo sapiens Integrin beta-2 Proteins 0.000 claims description 4
- 101000980823 Homo sapiens Leukocyte surface antigen CD53 Proteins 0.000 claims description 4
- 101001138062 Homo sapiens Leukocyte-associated immunoglobulin-like receptor 1 Proteins 0.000 claims description 4
- 101000608935 Homo sapiens Leukosialin Proteins 0.000 claims description 4
- 101000878605 Homo sapiens Low affinity immunoglobulin epsilon Fc receptor Proteins 0.000 claims description 4
- 101000984710 Homo sapiens Lymphocyte-specific protein 1 Proteins 0.000 claims description 4
- 101001051291 Homo sapiens Lysosomal-associated transmembrane protein 5 Proteins 0.000 claims description 4
- 101001024704 Homo sapiens Nck-associated protein 1-like Proteins 0.000 claims description 4
- 101000616502 Homo sapiens Phosphatidylinositol 3,4,5-trisphosphate 5-phosphatase 1 Proteins 0.000 claims description 4
- 101000935642 Homo sapiens Phosphoinositide 3-kinase adapter protein 1 Proteins 0.000 claims description 4
- 101000596046 Homo sapiens Plastin-2 Proteins 0.000 claims description 4
- 101001048969 Homo sapiens Protein FAM78A Proteins 0.000 claims description 4
- 101001051767 Homo sapiens Protein kinase C beta type Proteins 0.000 claims description 4
- 101000775749 Homo sapiens Proto-oncogene vav Proteins 0.000 claims description 4
- 101000712969 Homo sapiens Ras association domain-containing protein 5 Proteins 0.000 claims description 4
- 101001110313 Homo sapiens Ras-related C3 botulinum toxin substrate 2 Proteins 0.000 claims description 4
- 101000580036 Homo sapiens Ras-specific guanine nucleotide-releasing factor RalGPS2 Proteins 0.000 claims description 4
- 101000856696 Homo sapiens Rho GDP-dissociation inhibitor 2 Proteins 0.000 claims description 4
- 101001075565 Homo sapiens Rho GTPase-activating protein 30 Proteins 0.000 claims description 4
- 101000581151 Homo sapiens Rho GTPase-activating protein 9 Proteins 0.000 claims description 4
- 101000693722 Homo sapiens SAM and SH3 domain-containing protein 3 Proteins 0.000 claims description 4
- 101001092917 Homo sapiens SAM domain-containing protein SAMSN-1 Proteins 0.000 claims description 4
- 101000880431 Homo sapiens Serine/threonine-protein kinase 4 Proteins 0.000 claims description 4
- 101000863882 Homo sapiens Sialic acid-binding Ig-like lectin 7 Proteins 0.000 claims description 4
- 101000687654 Homo sapiens Sorting nexin-20 Proteins 0.000 claims description 4
- 101000891084 Homo sapiens T-cell activation Rho GTPase-activating protein Proteins 0.000 claims description 4
- 101000762938 Homo sapiens TOX high mobility group box family member 4 Proteins 0.000 claims description 4
- 101000881764 Homo sapiens Transcription elongation factor 1 homolog Proteins 0.000 claims description 4
- 101000800807 Homo sapiens Tumor necrosis factor alpha-induced protein 8 Proteins 0.000 claims description 4
- 101000799200 Homo sapiens Tumor necrosis factor alpha-induced protein 8-like protein 2 Proteins 0.000 claims description 4
- 101000912503 Homo sapiens Tyrosine-protein kinase Fgr Proteins 0.000 claims description 4
- 101001135589 Homo sapiens Tyrosine-protein phosphatase non-receptor type 22 Proteins 0.000 claims description 4
- 101000617285 Homo sapiens Tyrosine-protein phosphatase non-receptor type 6 Proteins 0.000 claims description 4
- 101001000119 Homo sapiens Unconventional myosin-If Proteins 0.000 claims description 4
- 101001000116 Homo sapiens Unconventional myosin-Ig Proteins 0.000 claims description 4
- 101000997307 Homo sapiens Voltage-gated potassium channel subunit beta-2 Proteins 0.000 claims description 4
- 102100025958 IGF-like family receptor 1 Human genes 0.000 claims description 4
- 102100022339 Integrin alpha-L Human genes 0.000 claims description 4
- 102100025390 Integrin beta-2 Human genes 0.000 claims description 4
- 108010064600 Intercellular Adhesion Molecule-3 Proteins 0.000 claims description 4
- 102100037871 Intercellular adhesion molecule 3 Human genes 0.000 claims description 4
- 102000000585 Interleukin-9 Human genes 0.000 claims description 4
- 108010002335 Interleukin-9 Proteins 0.000 claims description 4
- 102100024221 Leukocyte surface antigen CD53 Human genes 0.000 claims description 4
- 102100020943 Leukocyte-associated immunoglobulin-like receptor 1 Human genes 0.000 claims description 4
- 102100038007 Low affinity immunoglobulin epsilon Fc receptor Human genes 0.000 claims description 4
- 102100027105 Lymphocyte-specific protein 1 Human genes 0.000 claims description 4
- 102100024625 Lysosomal-associated transmembrane protein 5 Human genes 0.000 claims description 4
- 102100036942 Nck-associated protein 1-like Human genes 0.000 claims description 4
- 102100021797 Phosphatidylinositol 3,4,5-trisphosphate 5-phosphatase 1 Human genes 0.000 claims description 4
- 102100028238 Phosphoinositide 3-kinase adapter protein 1 Human genes 0.000 claims description 4
- 102100023831 Protein FAM78A Human genes 0.000 claims description 4
- 102100024923 Protein kinase C beta type Human genes 0.000 claims description 4
- 102100032190 Proto-oncogene vav Human genes 0.000 claims description 4
- 102100033239 Ras association domain-containing protein 5 Human genes 0.000 claims description 4
- 102100022129 Ras-related C3 botulinum toxin substrate 2 Human genes 0.000 claims description 4
- 102100027535 Ras-specific guanine nucleotide-releasing factor RalGPS2 Human genes 0.000 claims description 4
- 102100025622 Rho GDP-dissociation inhibitor 2 Human genes 0.000 claims description 4
- 102100027660 Rho GTPase-activating protein 15 Human genes 0.000 claims description 4
- 102100020887 Rho GTPase-activating protein 30 Human genes 0.000 claims description 4
- 102100027658 Rho GTPase-activating protein 9 Human genes 0.000 claims description 4
- 102100025544 SAM and SH3 domain-containing protein 3 Human genes 0.000 claims description 4
- 102100036195 SAM domain-containing protein SAMSN-1 Human genes 0.000 claims description 4
- 101150058731 STAT5A gene Proteins 0.000 claims description 4
- 102100037629 Serine/threonine-protein kinase 4 Human genes 0.000 claims description 4
- 102100029946 Sialic acid-binding Ig-like lectin 7 Human genes 0.000 claims description 4
- 102100024481 Signal transducer and activator of transcription 5A Human genes 0.000 claims description 4
- 102100024801 Sorting nexin-20 Human genes 0.000 claims description 4
- 102100040346 T-cell activation Rho GTPase-activating protein Human genes 0.000 claims description 4
- 102100026749 TOX high mobility group box family member 4 Human genes 0.000 claims description 4
- 102100037116 Transcription elongation factor 1 homolog Human genes 0.000 claims description 4
- 102100033649 Tumor necrosis factor alpha-induced protein 8 Human genes 0.000 claims description 4
- 102100034131 Tumor necrosis factor alpha-induced protein 8-like protein 2 Human genes 0.000 claims description 4
- 102100026150 Tyrosine-protein kinase Fgr Human genes 0.000 claims description 4
- 102100033138 Tyrosine-protein phosphatase non-receptor type 22 Human genes 0.000 claims description 4
- 102100021657 Tyrosine-protein phosphatase non-receptor type 6 Human genes 0.000 claims description 4
- 102100035825 Unconventional myosin-If Human genes 0.000 claims description 4
- 102100035824 Unconventional myosin-Ig Human genes 0.000 claims description 4
- 102100034074 Voltage-gated potassium channel subunit beta-2 Human genes 0.000 claims description 4
- 210000000066 myeloid cell Anatomy 0.000 claims description 4
- 230000036210 malignancy Effects 0.000 claims description 2
- 102100027816 Cytotoxic and regulatory T-cell molecule Human genes 0.000 claims 9
- 101001027081 Homo sapiens Killer cell immunoglobulin-like receptor 2DL1 Proteins 0.000 claims 9
- 101000831007 Homo sapiens T-cell immunoreceptor with Ig and ITIM domains Proteins 0.000 claims 9
- 101000946833 Homo sapiens T-cell surface glycoprotein CD8 beta chain Proteins 0.000 claims 9
- 102100037363 Killer cell immunoglobulin-like receptor 2DL1 Human genes 0.000 claims 9
- 108010015499 Protein Kinase C-theta Proteins 0.000 claims 9
- 102100021566 Protein kinase C theta type Human genes 0.000 claims 9
- 108010011033 Signaling Lymphocytic Activation Molecule Associated Protein Proteins 0.000 claims 9
- 102000013970 Signaling Lymphocytic Activation Molecule Associated Protein Human genes 0.000 claims 9
- 102100024834 T-cell immunoreceptor with Ig and ITIM domains Human genes 0.000 claims 9
- 102100034928 T-cell surface glycoprotein CD8 beta chain Human genes 0.000 claims 9
- 108010072917 class-I restricted T cell-associated molecule Proteins 0.000 claims 9
- 102100037077 Complement C1q subcomponent subunit A Human genes 0.000 claims 8
- 102100037799 DNA-binding protein Ikaros Human genes 0.000 claims 8
- 101000740726 Homo sapiens Complement C1q subcomponent subunit A Proteins 0.000 claims 8
- 101000599038 Homo sapiens DNA-binding protein Ikaros Proteins 0.000 claims 8
- 102100039498 Cytotoxic T-lymphocyte protein 4 Human genes 0.000 claims 6
- 102100029722 Ectonucleoside triphosphate diphosphohydrolase 1 Human genes 0.000 claims 6
- 102100030751 Eomesodermin homolog Human genes 0.000 claims 6
- 102100031508 Fc receptor-like protein 6 Human genes 0.000 claims 6
- 101150093535 Fcrl6 gene Proteins 0.000 claims 6
- 102100038395 Granzyme K Human genes 0.000 claims 6
- 108010007707 Hepatitis A Virus Cellular Receptor 2 Proteins 0.000 claims 6
- 102100034458 Hepatitis A virus cellular receptor 2 Human genes 0.000 claims 6
- 101000889276 Homo sapiens Cytotoxic T-lymphocyte protein 4 Proteins 0.000 claims 6
- 101001012447 Homo sapiens Ectonucleoside triphosphate diphosphohydrolase 1 Proteins 0.000 claims 6
- 101001064167 Homo sapiens Eomesodermin homolog Proteins 0.000 claims 6
- 101001033007 Homo sapiens Granzyme K Proteins 0.000 claims 6
- 101001046687 Homo sapiens Integrin alpha-E Proteins 0.000 claims 6
- 101001043809 Homo sapiens Interleukin-7 receptor subunit alpha Proteins 0.000 claims 6
- 101000972291 Homo sapiens Lymphoid enhancer-binding factor 1 Proteins 0.000 claims 6
- 101000835300 Homo sapiens Protein THEMIS Proteins 0.000 claims 6
- 101000679851 Homo sapiens Tumor necrosis factor receptor superfamily member 4 Proteins 0.000 claims 6
- 101000851370 Homo sapiens Tumor necrosis factor receptor superfamily member 9 Proteins 0.000 claims 6
- 101000818543 Homo sapiens Tyrosine-protein kinase ZAP-70 Proteins 0.000 claims 6
- 102100022341 Integrin alpha-E Human genes 0.000 claims 6
- 102100021593 Interleukin-7 receptor subunit alpha Human genes 0.000 claims 6
- 102100022699 Lymphoid enhancer-binding factor 1 Human genes 0.000 claims 6
- 102100034400 Nuclear factor of activated T-cells, cytoplasmic 2 Human genes 0.000 claims 6
- 102100026884 Pro-interleukin-16 Human genes 0.000 claims 6
- 102100026111 Protein THEMIS Human genes 0.000 claims 6
- 102100022153 Tumor necrosis factor receptor superfamily member 4 Human genes 0.000 claims 6
- 102100036856 Tumor necrosis factor receptor superfamily member 9 Human genes 0.000 claims 6
- 102100021125 Tyrosine-protein kinase ZAP-70 Human genes 0.000 claims 6
- 102100032937 CD40 ligand Human genes 0.000 claims 4
- 101000868215 Homo sapiens CD40 ligand Proteins 0.000 claims 4
- BSFODEXXVBBYOC-UHFFFAOYSA-N 8-[4-(dimethylamino)butan-2-ylamino]quinolin-6-ol Chemical compound C1=CN=C2C(NC(CCN(C)C)C)=CC(O)=CC2=C1 BSFODEXXVBBYOC-UHFFFAOYSA-N 0.000 claims 3
- 102100027399 A disintegrin and metalloproteinase with thrombospondin motifs 2 Human genes 0.000 claims 3
- 108091005662 ADAMTS2 Proteins 0.000 claims 3
- 102100031585 ADP-ribosyl cyclase/cyclic ADP-ribose hydrolase 1 Human genes 0.000 claims 3
- 108010004483 APOBEC-3G Deaminase Proteins 0.000 claims 3
- 102100022907 Acrosin-binding protein Human genes 0.000 claims 3
- 102100036732 Actin, aortic smooth muscle Human genes 0.000 claims 3
- 102100026402 Adhesion G protein-coupled receptor E2 Human genes 0.000 claims 3
- 102100022014 Angiopoietin-1 receptor Human genes 0.000 claims 3
- 102100034608 Angiopoietin-2 Human genes 0.000 claims 3
- 102100027836 Annexin-2 receptor Human genes 0.000 claims 3
- 102100029459 Apelin Human genes 0.000 claims 3
- 102100028218 Arf-GAP with coiled-coil, ANK repeat and PH domain-containing protein 1 Human genes 0.000 claims 3
- 241000501754 Astronotus ocellatus Species 0.000 claims 3
- 102100036301 C-C chemokine receptor type 7 Human genes 0.000 claims 3
- 102100036305 C-C chemokine receptor type 8 Human genes 0.000 claims 3
- 102100034673 C-C motif chemokine 3-like 1 Human genes 0.000 claims 3
- 102100021984 C-C motif chemokine 4-like Human genes 0.000 claims 3
- 102100028990 C-X-C chemokine receptor type 3 Human genes 0.000 claims 3
- 102100036189 C-X-C motif chemokine 3 Human genes 0.000 claims 3
- 102100032556 C-type lectin domain family 14 member A Human genes 0.000 claims 3
- 108010017009 CD11b Antigen Proteins 0.000 claims 3
- 102100021992 CD209 antigen Human genes 0.000 claims 3
- 102100029761 Cadherin-5 Human genes 0.000 claims 3
- 102100024965 Caspase recruitment domain-containing protein 11 Human genes 0.000 claims 3
- 102100033601 Collagen alpha-1(I) chain Human genes 0.000 claims 3
- 102100031611 Collagen alpha-1(III) chain Human genes 0.000 claims 3
- 102100022145 Collagen alpha-1(IV) chain Human genes 0.000 claims 3
- 102100031457 Collagen alpha-1(V) chain Human genes 0.000 claims 3
- 102100031519 Collagen alpha-1(VI) chain Human genes 0.000 claims 3
- 102100028257 Collagen alpha-1(XVI) chain Human genes 0.000 claims 3
- 102100036213 Collagen alpha-2(I) chain Human genes 0.000 claims 3
- 102100031518 Collagen alpha-2(VI) chain Human genes 0.000 claims 3
- 102100024338 Collagen alpha-3(VI) chain Human genes 0.000 claims 3
- 102100038076 DNA dC->dU-editing enzyme APOBEC-3G Human genes 0.000 claims 3
- 102100029790 Defensin-6 Human genes 0.000 claims 3
- 102100023471 E-selectin Human genes 0.000 claims 3
- 102100032064 EMILIN-2 Human genes 0.000 claims 3
- 102100027259 Ena/VASP-like protein Human genes 0.000 claims 3
- 102100037241 Endoglin Human genes 0.000 claims 3
- 102100038566 Endomucin Human genes 0.000 claims 3
- 102100038083 Endosialin Human genes 0.000 claims 3
- 102100038591 Endothelial cell-selective adhesion molecule Human genes 0.000 claims 3
- 102100031759 Endothelial cell-specific chemotaxis regulator Human genes 0.000 claims 3
- 102100021860 Endothelial cell-specific molecule 1 Human genes 0.000 claims 3
- 102100040351 FK506-binding protein 15 Human genes 0.000 claims 3
- 102100031509 Fibrillin-1 Human genes 0.000 claims 3
- 102100035290 Fibroblast growth factor 13 Human genes 0.000 claims 3
- 108090000379 Fibroblast growth factor 2 Proteins 0.000 claims 3
- 102100023589 Fibroblast growth factor-binding protein 2 Human genes 0.000 claims 3
- 102100038647 Fibroleukin Human genes 0.000 claims 3
- 102100031813 Fibulin-2 Human genes 0.000 claims 3
- 102100027581 Forkhead box protein P3 Human genes 0.000 claims 3
- 102100022629 Fructose-2,6-bisphosphatase Human genes 0.000 claims 3
- 102000054184 GADD45 Human genes 0.000 claims 3
- 102100022086 GRB2-related adapter protein 2 Human genes 0.000 claims 3
- 101710088083 Glomulin Proteins 0.000 claims 3
- 102100030386 Granzyme A Human genes 0.000 claims 3
- 102100030385 Granzyme B Human genes 0.000 claims 3
- 102100038393 Granzyme H Human genes 0.000 claims 3
- 102100035960 Hedgehog-interacting protein Human genes 0.000 claims 3
- 101710164669 Hedgehog-interacting protein Proteins 0.000 claims 3
- 102100029236 Hexokinase-3 Human genes 0.000 claims 3
- 101000777636 Homo sapiens ADP-ribosyl cyclase/cyclic ADP-ribose hydrolase 1 Proteins 0.000 claims 3
- 101000756551 Homo sapiens Acrosin-binding protein Proteins 0.000 claims 3
- 101000929319 Homo sapiens Actin, aortic smooth muscle Proteins 0.000 claims 3
- 101000594506 Homo sapiens Acyl-coenzyme A diphosphatase NUDT19 Proteins 0.000 claims 3
- 101000718211 Homo sapiens Adhesion G protein-coupled receptor E2 Proteins 0.000 claims 3
- 101000753291 Homo sapiens Angiopoietin-1 receptor Proteins 0.000 claims 3
- 101000924533 Homo sapiens Angiopoietin-2 Proteins 0.000 claims 3
- 101000698108 Homo sapiens Annexin-2 receptor Proteins 0.000 claims 3
- 101000771523 Homo sapiens Apelin Proteins 0.000 claims 3
- 101000724276 Homo sapiens Arf-GAP with coiled-coil, ANK repeat and PH domain-containing protein 1 Proteins 0.000 claims 3
- 101000716065 Homo sapiens C-C chemokine receptor type 7 Proteins 0.000 claims 3
- 101000716063 Homo sapiens C-C chemokine receptor type 8 Proteins 0.000 claims 3
- 101000946370 Homo sapiens C-C motif chemokine 3-like 1 Proteins 0.000 claims 3
- 101000896959 Homo sapiens C-C motif chemokine 4-like Proteins 0.000 claims 3
- 101000916050 Homo sapiens C-X-C chemokine receptor type 3 Proteins 0.000 claims 3
- 101000947193 Homo sapiens C-X-C motif chemokine 3 Proteins 0.000 claims 3
- 101000942280 Homo sapiens C-type lectin domain family 14 member A Proteins 0.000 claims 3
- 101000897416 Homo sapiens CD209 antigen Proteins 0.000 claims 3
- 101000794587 Homo sapiens Cadherin-5 Proteins 0.000 claims 3
- 101000761179 Homo sapiens Caspase recruitment domain-containing protein 11 Proteins 0.000 claims 3
- 101000993285 Homo sapiens Collagen alpha-1(III) chain Proteins 0.000 claims 3
- 101000901150 Homo sapiens Collagen alpha-1(IV) chain Proteins 0.000 claims 3
- 101000941708 Homo sapiens Collagen alpha-1(V) chain Proteins 0.000 claims 3
- 101000941581 Homo sapiens Collagen alpha-1(VI) chain Proteins 0.000 claims 3
- 101000860648 Homo sapiens Collagen alpha-1(XVI) chain Proteins 0.000 claims 3
- 101000875067 Homo sapiens Collagen alpha-2(I) chain Proteins 0.000 claims 3
- 101000941585 Homo sapiens Collagen alpha-2(VI) chain Proteins 0.000 claims 3
- 101000909506 Homo sapiens Collagen alpha-3(VI) chain Proteins 0.000 claims 3
- 101000865479 Homo sapiens Defensin-6 Proteins 0.000 claims 3
- 101000830440 Homo sapiens Differentially expressed in FDCP 6 homolog Proteins 0.000 claims 3
- 101000622123 Homo sapiens E-selectin Proteins 0.000 claims 3
- 101000921278 Homo sapiens EMILIN-2 Proteins 0.000 claims 3
- 101001057143 Homo sapiens Ena/VASP-like protein Proteins 0.000 claims 3
- 101000881679 Homo sapiens Endoglin Proteins 0.000 claims 3
- 101001030622 Homo sapiens Endomucin Proteins 0.000 claims 3
- 101000884275 Homo sapiens Endosialin Proteins 0.000 claims 3
- 101000882622 Homo sapiens Endothelial cell-selective adhesion molecule Proteins 0.000 claims 3
- 101000866525 Homo sapiens Endothelial cell-specific chemotaxis regulator Proteins 0.000 claims 3
- 101000897959 Homo sapiens Endothelial cell-specific molecule 1 Proteins 0.000 claims 3
- 101000891018 Homo sapiens FK506-binding protein 15 Proteins 0.000 claims 3
- 101000846893 Homo sapiens Fibrillin-1 Proteins 0.000 claims 3
- 101000827770 Homo sapiens Fibroblast growth factor-binding protein 2 Proteins 0.000 claims 3
- 101001031613 Homo sapiens Fibroleukin Proteins 0.000 claims 3
- 101001065274 Homo sapiens Fibulin-2 Proteins 0.000 claims 3
- 101000861452 Homo sapiens Forkhead box protein P3 Proteins 0.000 claims 3
- 101000823463 Homo sapiens Fructose-2,6-bisphosphatase Proteins 0.000 claims 3
- 101000900690 Homo sapiens GRB2-related adapter protein 2 Proteins 0.000 claims 3
- 101001009599 Homo sapiens Granzyme A Proteins 0.000 claims 3
- 101001009603 Homo sapiens Granzyme B Proteins 0.000 claims 3
- 101001033000 Homo sapiens Granzyme H Proteins 0.000 claims 3
- 101001066163 Homo sapiens Growth arrest and DNA damage-inducible protein GADD45 gamma Proteins 0.000 claims 3
- 101000840561 Homo sapiens Hexokinase-3 Proteins 0.000 claims 3
- 101001050472 Homo sapiens Integral membrane protein 2A Proteins 0.000 claims 3
- 101000852992 Homo sapiens Interleukin-12 subunit beta Proteins 0.000 claims 3
- 101001003140 Homo sapiens Interleukin-15 receptor subunit alpha Proteins 0.000 claims 3
- 101000971533 Homo sapiens Killer cell lectin-like receptor subfamily G member 1 Proteins 0.000 claims 3
- 101000941865 Homo sapiens Leucine-rich repeat neuronal protein 3 Proteins 0.000 claims 3
- 101000984200 Homo sapiens Leukocyte immunoglobulin-like receptor subfamily A member 3 Proteins 0.000 claims 3
- 101000984196 Homo sapiens Leukocyte immunoglobulin-like receptor subfamily A member 5 Proteins 0.000 claims 3
- 101000984206 Homo sapiens Leukocyte immunoglobulin-like receptor subfamily A member 6 Proteins 0.000 claims 3
- 101000984189 Homo sapiens Leukocyte immunoglobulin-like receptor subfamily B member 2 Proteins 0.000 claims 3
- 101000984192 Homo sapiens Leukocyte immunoglobulin-like receptor subfamily B member 3 Proteins 0.000 claims 3
- 101000984186 Homo sapiens Leukocyte immunoglobulin-like receptor subfamily B member 4 Proteins 0.000 claims 3
- 101000764294 Homo sapiens Lymphotoxin-beta Proteins 0.000 claims 3
- 101001004953 Homo sapiens Lysosomal acid lipase/cholesteryl ester hydrolase Proteins 0.000 claims 3
- 101001043321 Homo sapiens Lysyl oxidase homolog 1 Proteins 0.000 claims 3
- 101001011896 Homo sapiens Matrix metalloproteinase-19 Proteins 0.000 claims 3
- 101001014567 Homo sapiens Membrane-spanning 4-domains subfamily A member 7 Proteins 0.000 claims 3
- 101000947695 Homo sapiens Microfibrillar-associated protein 5 Proteins 0.000 claims 3
- 101001114673 Homo sapiens Multimerin-1 Proteins 0.000 claims 3
- 101001114675 Homo sapiens Multimerin-2 Proteins 0.000 claims 3
- 101001109518 Homo sapiens N-acetylneuraminate lyase Proteins 0.000 claims 3
- 101000829761 Homo sapiens N-arachidonyl glycine receptor Proteins 0.000 claims 3
- 101001109508 Homo sapiens NKG2-A/NKG2-B type II integral membrane protein Proteins 0.000 claims 3
- 101001109472 Homo sapiens NKG2-F type II integral membrane protein Proteins 0.000 claims 3
- 101001128156 Homo sapiens Nanos homolog 3 Proteins 0.000 claims 3
- 101000632154 Homo sapiens Ninjurin-1 Proteins 0.000 claims 3
- 101001124309 Homo sapiens Nitric oxide synthase, endothelial Proteins 0.000 claims 3
- 101000972834 Homo sapiens Normal mucosa of esophagus-specific gene 1 protein Proteins 0.000 claims 3
- 101000992377 Homo sapiens Osteoclast-associated immunoglobulin-like receptor Proteins 0.000 claims 3
- 101000614405 Homo sapiens P2X purinoceptor 1 Proteins 0.000 claims 3
- 101000609532 Homo sapiens Phosphoinositide-3-kinase-interacting protein 1 Proteins 0.000 claims 3
- 101001116302 Homo sapiens Platelet endothelial cell adhesion molecule Proteins 0.000 claims 3
- 101001126417 Homo sapiens Platelet-derived growth factor receptor alpha Proteins 0.000 claims 3
- 101000582936 Homo sapiens Pleckstrin Proteins 0.000 claims 3
- 101000612134 Homo sapiens Procollagen C-endopeptidase enhancer 1 Proteins 0.000 claims 3
- 101000728208 Homo sapiens Protein Aster-A Proteins 0.000 claims 3
- 101000876829 Homo sapiens Protein C-ets-1 Proteins 0.000 claims 3
- 101000866633 Homo sapiens Protein Hook homolog 3 Proteins 0.000 claims 3
- 101000781361 Homo sapiens Protein XRP2 Proteins 0.000 claims 3
- 101001026854 Homo sapiens Protein kinase C delta type Proteins 0.000 claims 3
- 101000995264 Homo sapiens Protein kinase C-binding protein NELL2 Proteins 0.000 claims 3
- 101000613620 Homo sapiens Protein mono-ADP-ribosyltransferase PARP15 Proteins 0.000 claims 3
- 101001138006 Homo sapiens Purine nucleoside phosphorylase LACC1 Proteins 0.000 claims 3
- 101001061893 Homo sapiens RAS protein activator like-3 Proteins 0.000 claims 3
- 101001130471 Homo sapiens Ras-interacting protein 1 Proteins 0.000 claims 3
- 101001099871 Homo sapiens Ras-related protein Rab-42 Proteins 0.000 claims 3
- 101000738772 Homo sapiens Receptor-type tyrosine-protein phosphatase beta Proteins 0.000 claims 3
- 101001081189 Homo sapiens Rho GTPase-activating protein 45 Proteins 0.000 claims 3
- 101000692943 Homo sapiens Ribonuclease K6 Proteins 0.000 claims 3
- 101000650590 Homo sapiens Roundabout homolog 4 Proteins 0.000 claims 3
- 101000633784 Homo sapiens SLAM family member 7 Proteins 0.000 claims 3
- 101000836983 Homo sapiens Secretoglobin family 1D member 1 Proteins 0.000 claims 3
- 101000611251 Homo sapiens Serine/threonine-protein phosphatase 2B catalytic subunit gamma isoform Proteins 0.000 claims 3
- 101000709473 Homo sapiens Sialic acid-binding Ig-like lectin 14 Proteins 0.000 claims 3
- 101000663635 Homo sapiens Sphingosine kinase 1 Proteins 0.000 claims 3
- 101000633700 Homo sapiens Src kinase-associated phosphoprotein 1 Proteins 0.000 claims 3
- 101000832225 Homo sapiens Stabilin-1 Proteins 0.000 claims 3
- 101000706156 Homo sapiens Syntaxin-11 Proteins 0.000 claims 3
- 101000596234 Homo sapiens T-cell surface protein tactile Proteins 0.000 claims 3
- 101000663002 Homo sapiens TNFAIP3-interacting protein 3 Proteins 0.000 claims 3
- 101000633605 Homo sapiens Thrombospondin-2 Proteins 0.000 claims 3
- 101000800116 Homo sapiens Thy-1 membrane glycoprotein Proteins 0.000 claims 3
- 101000979190 Homo sapiens Transcription factor MafB Proteins 0.000 claims 3
- 101000652736 Homo sapiens Transgelin Proteins 0.000 claims 3
- 101000851627 Homo sapiens Transmembrane channel-like protein 6 Proteins 0.000 claims 3
- 101000904724 Homo sapiens Transmembrane glycoprotein NMB Proteins 0.000 claims 3
- 101000611183 Homo sapiens Tumor necrosis factor Proteins 0.000 claims 3
- 101000830600 Homo sapiens Tumor necrosis factor ligand superfamily member 13 Proteins 0.000 claims 3
- 101000638161 Homo sapiens Tumor necrosis factor ligand superfamily member 6 Proteins 0.000 claims 3
- 101000801234 Homo sapiens Tumor necrosis factor receptor superfamily member 18 Proteins 0.000 claims 3
- 101000851376 Homo sapiens Tumor necrosis factor receptor superfamily member 8 Proteins 0.000 claims 3
- 101000847156 Homo sapiens Tumor necrosis factor-inducible gene 6 protein Proteins 0.000 claims 3
- 101000753253 Homo sapiens Tyrosine-protein kinase receptor Tie-1 Proteins 0.000 claims 3
- 101000742596 Homo sapiens Vascular endothelial growth factor C Proteins 0.000 claims 3
- 101000851018 Homo sapiens Vascular endothelial growth factor receptor 1 Proteins 0.000 claims 3
- 101000976590 Homo sapiens Zinc finger protein 101 Proteins 0.000 claims 3
- 101000599046 Homo sapiens Zinc finger protein Eos Proteins 0.000 claims 3
- 101000599037 Homo sapiens Zinc finger protein Helios Proteins 0.000 claims 3
- 101150082255 IGSF6 gene Proteins 0.000 claims 3
- 102100022532 Immunoglobulin superfamily member 6 Human genes 0.000 claims 3
- 102100023351 Integral membrane protein 2A Human genes 0.000 claims 3
- 102100022338 Integrin alpha-M Human genes 0.000 claims 3
- 102100036701 Interleukin-12 subunit beta Human genes 0.000 claims 3
- 102100020789 Interleukin-15 receptor subunit alpha Human genes 0.000 claims 3
- 102100033096 Interleukin-17D Human genes 0.000 claims 3
- 108010017411 Interleukin-21 Receptors Proteins 0.000 claims 3
- 102100030699 Interleukin-21 receptor Human genes 0.000 claims 3
- 108010066979 Interleukin-27 Proteins 0.000 claims 3
- 102100021457 Killer cell lectin-like receptor subfamily G member 1 Human genes 0.000 claims 3
- 102100037691 Kinesin-like protein KIF20B Human genes 0.000 claims 3
- 108050007394 Kinesin-like protein KIF20B Proteins 0.000 claims 3
- 102100032657 Leucine-rich repeat neuronal protein 3 Human genes 0.000 claims 3
- 102100025556 Leukocyte immunoglobulin-like receptor subfamily A member 3 Human genes 0.000 claims 3
- 102100025574 Leukocyte immunoglobulin-like receptor subfamily A member 5 Human genes 0.000 claims 3
- 102100025553 Leukocyte immunoglobulin-like receptor subfamily A member 6 Human genes 0.000 claims 3
- 102100025583 Leukocyte immunoglobulin-like receptor subfamily B member 2 Human genes 0.000 claims 3
- 102100025582 Leukocyte immunoglobulin-like receptor subfamily B member 3 Human genes 0.000 claims 3
- 102100025578 Leukocyte immunoglobulin-like receptor subfamily B member 4 Human genes 0.000 claims 3
- 102100026894 Lymphotoxin-beta Human genes 0.000 claims 3
- 102100026001 Lysosomal acid lipase/cholesteryl ester hydrolase Human genes 0.000 claims 3
- 102100021958 Lysyl oxidase homolog 1 Human genes 0.000 claims 3
- 102100030218 Matrix metalloproteinase-19 Human genes 0.000 claims 3
- 102100036203 Microfibrillar-associated protein 5 Human genes 0.000 claims 3
- 102100023354 Multimerin-1 Human genes 0.000 claims 3
- 102100023346 Multimerin-2 Human genes 0.000 claims 3
- 102100022686 N-acetylneuraminate lyase Human genes 0.000 claims 3
- 102100023414 N-arachidonyl glycine receptor Human genes 0.000 claims 3
- 102100022682 NKG2-A/NKG2-B type II integral membrane protein Human genes 0.000 claims 3
- 102100022700 NKG2-F type II integral membrane protein Human genes 0.000 claims 3
- 102100031893 Nanos homolog 3 Human genes 0.000 claims 3
- 102100027894 Ninjurin-1 Human genes 0.000 claims 3
- 102100022646 Normal mucosa of esophagus-specific gene 1 protein Human genes 0.000 claims 3
- 102100032159 Osteoclast-associated immunoglobulin-like receptor Human genes 0.000 claims 3
- 102100040444 P2X purinoceptor 1 Human genes 0.000 claims 3
- 102100024894 PR domain zinc finger protein 1 Human genes 0.000 claims 3
- 102100039472 Phosphoinositide-3-kinase-interacting protein 1 Human genes 0.000 claims 3
- 102100024616 Platelet endothelial cell adhesion molecule Human genes 0.000 claims 3
- 108010051742 Platelet-Derived Growth Factor beta Receptor Proteins 0.000 claims 3
- 102100030485 Platelet-derived growth factor receptor alpha Human genes 0.000 claims 3
- 102100026547 Platelet-derived growth factor receptor beta Human genes 0.000 claims 3
- 102100030264 Pleckstrin Human genes 0.000 claims 3
- 108010009975 Positive Regulatory Domain I-Binding Factor 1 Proteins 0.000 claims 3
- 102100033874 Probable sodium-coupled neutral amino acid transporter 6 Human genes 0.000 claims 3
- 102100041026 Procollagen C-endopeptidase enhancer 1 Human genes 0.000 claims 3
- 102100023832 Prolyl endopeptidase FAP Human genes 0.000 claims 3
- 102100029800 Protein Aster-A Human genes 0.000 claims 3
- 102100035251 Protein C-ets-1 Human genes 0.000 claims 3
- 102100033154 Protein XRP2 Human genes 0.000 claims 3
- 102100037340 Protein kinase C delta type Human genes 0.000 claims 3
- 102100034433 Protein kinase C-binding protein NELL2 Human genes 0.000 claims 3
- 102100040846 Protein mono-ADP-ribosyltransferase PARP15 Human genes 0.000 claims 3
- 108010007100 Pulmonary Surfactant-Associated Protein A Proteins 0.000 claims 3
- 102100027773 Pulmonary surfactant-associated protein A2 Human genes 0.000 claims 3
- 102100020863 Purine nucleoside phosphorylase LACC1 Human genes 0.000 claims 3
- 102100029556 RAS protein activator like-3 Human genes 0.000 claims 3
- 102100031429 Ras-interacting protein 1 Human genes 0.000 claims 3
- 102100038506 Ras-related protein Rab-42 Human genes 0.000 claims 3
- 102100037424 Receptor-type tyrosine-protein phosphatase beta Human genes 0.000 claims 3
- 102100027748 Rho GTPase-activating protein 45 Human genes 0.000 claims 3
- 102100026386 Ribonuclease K6 Human genes 0.000 claims 3
- 102100027701 Roundabout homolog 4 Human genes 0.000 claims 3
- 102100029198 SLAM family member 7 Human genes 0.000 claims 3
- 108091006935 SLC38A6 Proteins 0.000 claims 3
- 101000733871 Saccharomyces cerevisiae (strain ATCC 204508 / S288c) 60S ribosomal protein L4-A Proteins 0.000 claims 3
- 101000733875 Saccharomyces cerevisiae (strain ATCC 204508 / S288c) 60S ribosomal protein L4-B Proteins 0.000 claims 3
- 102100040320 Serine/threonine-protein phosphatase 2B catalytic subunit gamma isoform Human genes 0.000 claims 3
- 102100034370 Sialic acid-binding Ig-like lectin 14 Human genes 0.000 claims 3
- 101150043341 Socs3 gene Proteins 0.000 claims 3
- 102100039024 Sphingosine kinase 1 Human genes 0.000 claims 3
- 102100029208 Src kinase-associated phosphoprotein 1 Human genes 0.000 claims 3
- 102100024471 Stabilin-1 Human genes 0.000 claims 3
- 102100032891 Superoxide dismutase [Mn], mitochondrial Human genes 0.000 claims 3
- 102000058015 Suppressor of Cytokine Signaling 3 Human genes 0.000 claims 3
- 108700027337 Suppressor of Cytokine Signaling 3 Proteins 0.000 claims 3
- 102100031115 Syntaxin-11 Human genes 0.000 claims 3
- 102100035268 T-cell surface protein tactile Human genes 0.000 claims 3
- 102100037666 TNFAIP3-interacting protein 3 Human genes 0.000 claims 3
- 102100029529 Thrombospondin-2 Human genes 0.000 claims 3
- 102100033523 Thy-1 membrane glycoprotein Human genes 0.000 claims 3
- 102100023234 Transcription factor MafB Human genes 0.000 claims 3
- 102100031013 Transgelin Human genes 0.000 claims 3
- 102100036810 Transmembrane channel-like protein 6 Human genes 0.000 claims 3
- 102100023935 Transmembrane glycoprotein NMB Human genes 0.000 claims 3
- 102100040247 Tumor necrosis factor Human genes 0.000 claims 3
- 102100024585 Tumor necrosis factor ligand superfamily member 13 Human genes 0.000 claims 3
- 102100031988 Tumor necrosis factor ligand superfamily member 6 Human genes 0.000 claims 3
- 102100033728 Tumor necrosis factor receptor superfamily member 18 Human genes 0.000 claims 3
- 102100036857 Tumor necrosis factor receptor superfamily member 8 Human genes 0.000 claims 3
- 102100032807 Tumor necrosis factor-inducible gene 6 protein Human genes 0.000 claims 3
- 102100022007 Tyrosine-protein kinase receptor Tie-1 Human genes 0.000 claims 3
- 101150045640 VWF gene Proteins 0.000 claims 3
- 108010053099 Vascular Endothelial Growth Factor Receptor-2 Proteins 0.000 claims 3
- 102100038232 Vascular endothelial growth factor C Human genes 0.000 claims 3
- 102100033178 Vascular endothelial growth factor receptor 1 Human genes 0.000 claims 3
- 102100033177 Vascular endothelial growth factor receptor 2 Human genes 0.000 claims 3
- 102100023576 Zinc finger protein 101 Human genes 0.000 claims 3
- 102100037793 Zinc finger protein Eos Human genes 0.000 claims 3
- 102100037796 Zinc finger protein Helios Human genes 0.000 claims 3
- 108010029483 alpha 1 Chain Collagen Type I Proteins 0.000 claims 3
- 108010045815 superoxide dismutase 2 Proteins 0.000 claims 3
- 102100036537 von Willebrand factor Human genes 0.000 claims 3
- 101001018100 Homo sapiens Lysozyme C Proteins 0.000 claims 2
- 101001054878 Homo sapiens Tyrosine-protein kinase Lyn Proteins 0.000 claims 2
- 102100033468 Lysozyme C Human genes 0.000 claims 2
- 102100026857 Tyrosine-protein kinase Lyn Human genes 0.000 claims 2
- 229920002477 rna polymer Polymers 0.000 description 99
- 238000005516 engineering process Methods 0.000 description 35
- 238000012163 sequencing technique Methods 0.000 description 21
- 238000003559 RNA-seq method Methods 0.000 description 19
- 238000010586 diagram Methods 0.000 description 14
- 238000010801 machine learning Methods 0.000 description 14
- 102000039446 nucleic acids Human genes 0.000 description 13
- 108020004707 nucleic acids Proteins 0.000 description 13
- 150000007523 nucleic acids Chemical class 0.000 description 13
- 210000001519 tissue Anatomy 0.000 description 12
- 230000001413 cellular effect Effects 0.000 description 11
- 230000006870 function Effects 0.000 description 11
- 239000000523 sample Substances 0.000 description 11
- 230000000875 corresponding effect Effects 0.000 description 10
- 210000004881 tumor cell Anatomy 0.000 description 9
- 238000004458 analytical method Methods 0.000 description 8
- 238000007796 conventional method Methods 0.000 description 7
- 239000003550 marker Substances 0.000 description 7
- 102000053602 DNA Human genes 0.000 description 6
- 108020004414 DNA Proteins 0.000 description 6
- 210000002381 plasma Anatomy 0.000 description 6
- 238000002360 preparation method Methods 0.000 description 6
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 238000003908 quality control method Methods 0.000 description 4
- 238000007482 whole exome sequencing Methods 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 238000001574 biopsy Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000007481 next generation sequencing Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000011282 treatment Methods 0.000 description 3
- 238000012800 visualization Methods 0.000 description 3
- 108091028043 Nucleic acid sequence Proteins 0.000 description 2
- 230000004913 activation Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 210000004369 blood Anatomy 0.000 description 2
- 239000008280 blood Substances 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 238000000684 flow cytometry Methods 0.000 description 2
- 230000002068 genetic effect Effects 0.000 description 2
- 238000002955 isolation Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000035772 mutation Effects 0.000 description 2
- 239000002773 nucleotide Substances 0.000 description 2
- 125000003729 nucleotide group Chemical class 0.000 description 2
- 230000008506 pathogenesis Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000007480 sanger sequencing Methods 0.000 description 2
- 230000004083 survival effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 210000004981 tumor-associated macrophage Anatomy 0.000 description 2
- 238000012070 whole genome sequencing analysis Methods 0.000 description 2
- 206010006187 Breast cancer Diseases 0.000 description 1
- 208000026310 Breast neoplasm Diseases 0.000 description 1
- 102000017420 CD3 protein, epsilon/gamma/delta subunit Human genes 0.000 description 1
- 108050005493 CD3 protein, epsilon/gamma/delta subunit Proteins 0.000 description 1
- 210000004366 CD4-positive T-lymphocyte Anatomy 0.000 description 1
- 206010009944 Colon cancer Diseases 0.000 description 1
- 208000001333 Colorectal Neoplasms Diseases 0.000 description 1
- 102000004127 Cytokines Human genes 0.000 description 1
- 108090000695 Cytokines Proteins 0.000 description 1
- 206010061818 Disease progression Diseases 0.000 description 1
- 108700024394 Exon Proteins 0.000 description 1
- 102000010834 Extracellular Matrix Proteins Human genes 0.000 description 1
- 108010037362 Extracellular Matrix Proteins Proteins 0.000 description 1
- 208000034826 Genetic Predisposition to Disease Diseases 0.000 description 1
- 102000049772 Interleukin-16 Human genes 0.000 description 1
- 208000008839 Kidney Neoplasms Diseases 0.000 description 1
- 238000012313 Kruskal-Wallis test Methods 0.000 description 1
- 206010058467 Lung neoplasm malignant Diseases 0.000 description 1
- 210000004322 M2 macrophage Anatomy 0.000 description 1
- 102000002673 NFATC Transcription Factors Human genes 0.000 description 1
- 108091034117 Oligonucleotide Proteins 0.000 description 1
- 108700026244 Open Reading Frames Proteins 0.000 description 1
- 206010061902 Pancreatic neoplasm Diseases 0.000 description 1
- 206010060862 Prostate cancer Diseases 0.000 description 1
- 208000000236 Prostatic Neoplasms Diseases 0.000 description 1
- 108700005075 Regulator Genes Proteins 0.000 description 1
- 206010038389 Renal cancer Diseases 0.000 description 1
- 208000005718 Stomach Neoplasms Diseases 0.000 description 1
- JLCPHMBAVCMARE-UHFFFAOYSA-N [3-[[3-[[3-[[3-[[3-[[3-[[3-[[3-[[3-[[3-[[3-[[5-(2-amino-6-oxo-1H-purin-9-yl)-3-[[3-[[3-[[3-[[3-[[3-[[5-(2-amino-6-oxo-1H-purin-9-yl)-3-[[5-(2-amino-6-oxo-1H-purin-9-yl)-3-hydroxyoxolan-2-yl]methoxy-hydroxyphosphoryl]oxyoxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(5-methyl-2,4-dioxopyrimidin-1-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(6-aminopurin-9-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(6-aminopurin-9-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(6-aminopurin-9-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(6-aminopurin-9-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxyoxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(5-methyl-2,4-dioxopyrimidin-1-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(4-amino-2-oxopyrimidin-1-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(5-methyl-2,4-dioxopyrimidin-1-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(5-methyl-2,4-dioxopyrimidin-1-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(6-aminopurin-9-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(6-aminopurin-9-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(4-amino-2-oxopyrimidin-1-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(4-amino-2-oxopyrimidin-1-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(4-amino-2-oxopyrimidin-1-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(6-aminopurin-9-yl)oxolan-2-yl]methoxy-hydroxyphosphoryl]oxy-5-(4-amino-2-oxopyrimidin-1-yl)oxolan-2-yl]methyl [5-(6-aminopurin-9-yl)-2-(hydroxymethyl)oxolan-3-yl] hydrogen phosphate Polymers Cc1cn(C2CC(OP(O)(=O)OCC3OC(CC3OP(O)(=O)OCC3OC(CC3O)n3cnc4c3nc(N)[nH]c4=O)n3cnc4c3nc(N)[nH]c4=O)C(COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3COP(O)(=O)OC3CC(OC3CO)n3cnc4c(N)ncnc34)n3ccc(N)nc3=O)n3cnc4c(N)ncnc34)n3ccc(N)nc3=O)n3ccc(N)nc3=O)n3ccc(N)nc3=O)n3cnc4c(N)ncnc34)n3cnc4c(N)ncnc34)n3cc(C)c(=O)[nH]c3=O)n3cc(C)c(=O)[nH]c3=O)n3ccc(N)nc3=O)n3cc(C)c(=O)[nH]c3=O)n3cnc4c3nc(N)[nH]c4=O)n3cnc4c(N)ncnc34)n3cnc4c(N)ncnc34)n3cnc4c(N)ncnc34)n3cnc4c(N)ncnc34)O2)c(=O)[nH]c1=O JLCPHMBAVCMARE-UHFFFAOYSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 239000011324 bead Substances 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 210000003850 cellular structure Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 210000003690 classically activated macrophage Anatomy 0.000 description 1
- 238000011109 contamination Methods 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 210000001151 cytotoxic T lymphocyte Anatomy 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 210000004443 dendritic cell Anatomy 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000009274 differential gene expression Effects 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 230000005750 disease progression Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 206010017758 gastric cancer Diseases 0.000 description 1
- 230000012010 growth Effects 0.000 description 1
- 239000003102 growth factor Substances 0.000 description 1
- 230000003394 haemopoietic effect Effects 0.000 description 1
- 201000010536 head and neck cancer Diseases 0.000 description 1
- 208000014829 head and neck neoplasm Diseases 0.000 description 1
- 210000002443 helper t lymphocyte Anatomy 0.000 description 1
- 230000037451 immune surveillance Effects 0.000 description 1
- 210000000987 immune system Anatomy 0.000 description 1
- 238000003364 immunohistochemistry Methods 0.000 description 1
- 238000000338 in vitro Methods 0.000 description 1
- 210000004263 induced pluripotent stem cell Anatomy 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- PGHMRUGBZOYCAA-ADZNBVRBSA-N ionomycin Chemical compound O1[C@H](C[C@H](O)[C@H](C)[C@H](O)[C@H](C)/C=C/C[C@@H](C)C[C@@H](C)C(/O)=C/C(=O)[C@@H](C)C[C@@H](C)C[C@@H](CCC(O)=O)C)CC[C@@]1(C)[C@@H]1O[C@](C)([C@@H](C)O)CC1 PGHMRUGBZOYCAA-ADZNBVRBSA-N 0.000 description 1
- PGHMRUGBZOYCAA-UHFFFAOYSA-N ionomycin Natural products O1C(CC(O)C(C)C(O)C(C)C=CCC(C)CC(C)C(O)=CC(=O)C(C)CC(C)CC(CCC(O)=O)C)CCC1(C)C1OC(C)(C(C)O)CC1 PGHMRUGBZOYCAA-UHFFFAOYSA-N 0.000 description 1
- 150000002500 ions Chemical class 0.000 description 1
- 201000010982 kidney cancer Diseases 0.000 description 1
- 238000013173 literature analysis Methods 0.000 description 1
- 201000005202 lung cancer Diseases 0.000 description 1
- 208000020816 lung neoplasm Diseases 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 208000015486 malignant pancreatic neoplasm Diseases 0.000 description 1
- 201000001441 melanoma Diseases 0.000 description 1
- 108020004999 messenger RNA Proteins 0.000 description 1
- 230000002503 metabolic effect Effects 0.000 description 1
- 238000012316 non-parametric ANOVA Methods 0.000 description 1
- 210000004882 non-tumor cell Anatomy 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 201000002528 pancreatic cancer Diseases 0.000 description 1
- 208000008443 pancreatic carcinoma Diseases 0.000 description 1
- PHEDXBVPIONUQT-RGYGYFBISA-N phorbol 13-acetate 12-myristate Chemical compound C([C@]1(O)C(=O)C(C)=C[C@H]1[C@@]1(O)[C@H](C)[C@H]2OC(=O)CCCCCCCCCCCCC)C(CO)=C[C@H]1[C@H]1[C@]2(OC(C)=O)C1(C)C PHEDXBVPIONUQT-RGYGYFBISA-N 0.000 description 1
- 230000035790 physiological processes and functions Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000004393 prognosis Methods 0.000 description 1
- 230000002829 reductive effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000000284 resting effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000011664 signaling Effects 0.000 description 1
- 210000004927 skin cell Anatomy 0.000 description 1
- 239000000243 solution Substances 0.000 description 1
- 201000011549 stomach cancer Diseases 0.000 description 1
- 238000002560 therapeutic procedure Methods 0.000 description 1
- 230000004797 therapeutic response Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 238000012049 whole transcriptome sequencing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G16B25/00—ICT specially adapted for hybridisation; ICT specially adapted for gene or protein expression
- G16B25/10—Gene or protein expression profiling; Expression-ratio estimation or normalisation
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G16B40/00—ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding
- G16B40/20—Supervised data analysis
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Definitions
- a tumor mass may comprise a population of malignant cells (e.g., cancer cells) and a microenvironment which may include, for example, immune cells, fibroblasts, and extracellular matrix proteins.
- malignant cells e.g., cancer cells
- microenvironment which may include, for example, immune cells, fibroblasts, and extracellular matrix proteins.
- Some embodiments provide for a method comprising using at least one computer hardware processor to perform: obtaining expression data for a biological sample, the biological sample previously obtained from a subject, the expression data including first expression data associated with a first set of genes associated with a first cell type; determining first a cell composition percentage for the first cell type using the expression data and one or more non-linear regression models including a first non-linear regression model, wherein the first cell composition percentage indicates an estimated percentage of cells of the first cell type in the biological sample, wherein determining the first cell composition percentage for the first cell type comprises: processing the first expression data with the first non-linear regression model to determine the first cell composition percentage for the first cell type; and outputting the first cell composition percentage.
- Some embodiments provide for a system, comprising: at least one hardware processor; and at least one non-transitory computer-readable storage medium storing processor-executable instructions that, when executed by the at least one hardware processor, cause the at least one hardware processor to perform: obtaining expression data for a biological sample, the biological sample previously obtained from a subject, the expression data including first expression data associated with a first set of genes associated with a first cell type; determining a first cell composition percentage for the first cell type using the expression data and one or more non-linear regression models including a first non-linear regression model, wherein the first cell composition percentage indicates an estimated percentage of cells of the first cell type in the biological sample, wherein determining the first cell composition percentage for the first cell type comprises: processing the first expression data with the first non-linear regression model to determine the first cell composition percentage for the first cell type; and outputting the first cell composition percentage.
- Some embodiments provide for at least one non-transitory computer-readable storage medium storing processor-executable instructions that, when executed by at least one hardware processor, cause the at least one hardware processor to perform: obtaining expression data for a biological sample, the biological sample previously obtained from a subject, the expression data including first expression data associated with a first set of genes associated with a first cell type; determining a first cell composition percentage for the first cell type using the expression data and one or more non-linear regression models including a first non-linear regression model, wherein the first cell composition percentage indicates an estimated percentage of cells of the first cell type in the biological sample, wherein determining the first cell composition percentage for the first cell type comprises: processing the first expression data with the first non-linear regression model to determine the first cell composition percentage for the first cell type; and outputting the first cell composition percentage.
- Some embodiments provide for a method comprising using at least one computer hardware processor to perform: obtaining RNA expression data for a biological sample, the biological sample previously obtained from a subject having, suspected of having, or at risk of having cancer, wherein the RNA expression data includes first RNA expression data associated with a first set of genes associated with a first cell type, wherein the first RNA expression data includes expression data for at least 10 genes selected from the group of genes for the first cell type in Table 2, wherein the first cell type is selected from the group consisting of B cells, CD4+ T cells, CD8+ T cells, endothelial cells, fibroblasts, lymphocytes, macrophages, monocytes, NK cells, neutrophils, and T cells; and determining a first cell composition percentage for the first cell type, using the first RNA expression data, the first cell composition percentage indicating an estimated percentage of cells of the first cell type in the biological sample, wherein determining the first cell composition percentage for the first cell type comprises: providing the first RNA expression data as input to a first non
- RNA expression data for a biological sample, the biological sample previously obtained from a subject having, suspected of having, or at risk of having cancer
- the RNA expression data includes first RNA expression data associated with a first set of genes associated with a first cell type
- the first RNA expression data includes expression data for at least 10 genes selected from the group of genes for the first cell type in Table 2, wherein the first cell type is selected from the group consisting of B cells, CD4+ T cells, CD8+ T cells, endothelial cells, fibroblasts, lymphocytes, macrophages, monocytes, NK cells, neutrophils, and T cells; and determining a first cell composition percentage for the first cell type, using the first RNA expression data, the first cell composition percentage indicating an estimated percentage of cells of the
- Some embodiments provide for at least one non-transitory computer-readable storage medium storing processor-executable instructions that, when executed by at least one hardware processor, cause the at least one hardware processor to perform: obtaining RNA expression data for a biological sample, the biological sample previously obtained from a subject having, suspected of having, or at risk of having cancer, wherein the RNA expression data includes first RNA expression data associated with a first set of genes associated with a first cell type, wherein the first RNA expression data includes expression data for at least 10 genes selected from the group of genes for the first cell type in Table 2, wherein the first cell type is selected from the group consisting of B cells, CD4+ T cells, CD8+ T cells, endothelial cells, fibroblasts, lymphocytes, macrophages, monocytes, NK cells, neutrophils, and T cells; and determining a first cell composition percentage for the first cell type, using the first RNA expression data, the first cell composition percentage indicating an estimated percentage of cells of the first cell type in the biological sample, wherein
- Some embodiments provide for a method comprising: using at least one computer hardware processor to perform: obtaining training data comprising simulated RNA expression data, the simulated RNA expression data including first RNA expression data for first genes associated with a first cell type and second RNA expression data for second genes associated with a second cell type different from the first cell type; and training a plurality of non-linear regression models to estimate percentages of RNA from one or more respective cell types, the plurality of non-linear regression models comprising a first non-linear regression model for estimating percentage of RNA from the first cell type and a second non-linear regression model for estimating percentage of RNA from the second cell type, wherein training the plurality of non-linear regression models comprises training the first non-linear regression model at least in part by: generating, using the first non-linear regression model and the first RNA expression data, an estimated percentage of RNA from the first cell type; and updating parameters of the first non-linear regression model using the estimated percentage of RNA from the first cell type; and outputting the
- Some embodiments provide for a system, comprising: at least one computer hardware processor; and at least one non-transitory computer-readable storage medium storing processor executable instructions that, when executed by the at least one computer hardware processor, cause the at least one computer hardware processor to perform: obtaining training data comprising simulated RNA expression data, the simulated RNA expression data including first RNA expression data for first genes associated with a first cell type and second RNA expression data for second genes associated with a second cell type different from the first cell type; and training a plurality of non-linear regression models to estimate percentages of RNA from one or more respective cell types, the plurality of non-linear regression models comprising a first non-linear regression model for estimating percentage of RNA from the first cell type and a second non-linear regression model for estimating percentage of RNA from the second cell type, wherein training the plurality of non-linear regression models comprises training the first non-linear regression model at least in part by: generating, using the first non-linear regression model and the first RNA expression data
- Some embodiments provide at least one non-transitory computer-readable storage medium storing processor executable instructions that, when executed by the at least one computer hardware processor, cause the at least one computer hardware processor to perform: obtaining training data comprising simulated RNA expression data, the simulated RNA expression data including first RNA expression data for first genes associated with a first cell type and second RNA expression data for second genes associated with a second cell type different from the first cell type; and training a plurality of non-linear regression models to estimate percentages of RNA from one or more respective cell types, the plurality of non-linear regression models comprising a first non-linear regression model for estimating percentage of RNA from the first cell type and a second non-linear regression model for estimating percentage of RNA from the second cell type, wherein training the plurality of non-linear regression models comprises training the first non-linear regression mode at least in part by: generating, using the first non-linear regression model and the first RNA expression data, an estimated percentage of RNA from the first cell type; and updating parameters
- Some embodiments provide for a method comprising using at least one computer hardware processor to perform: obtaining expression data for a biological sample, the biological sample previously obtained from a subject having, suspected of having, or at risk of having cancer; obtaining a plurality of expression profiles for a corresponding plurality of cell types, each of the expression profiles comprising respective expression data from one or more genes associated with a respective cell type from the plurality of cell types; and determining a plurality of cell composition percentages for the plurality of cell types at least in part by optimizing a piecewise continuous error function between the expression data and the plurality of expression profiles.
- Some embodiments provide for a system, comprising: at least one computer hardware processor; and at least one computer readable storage medium storing processor executable instructions that, when executed by the at least one computer hardware processor, cause the at least one computer hardware processor to perform obtaining expression data for a biological sample, the biological sample previously obtained from a subject having, suspected of having, or at risk of having cancer; obtaining a plurality of expression profiles for a corresponding plurality of cell types, each of the expression profiles comprising respective expression data from one or more genes associated with a respective cell type from the plurality of cell types; and determining a plurality of cell composition percentages for the plurality of cell types at least in part by optimizing a piecewise continuous error function between the expression data and the plurality of expression profiles.
- Some embodiments provide for at least one computer-readable storage medium storing processor executable instructions that, when executed by at least one computer hardware processor, cause the at least one computer hardware processor to perform: obtaining expression data for a biological sample, the biological sample previously obtained from a subject having, suspected of having, or at risk of having cancer; obtaining a plurality of expression profiles for a corresponding plurality of cell types, each of the expression profiles comprising respective expression data from one or more genes associated with a respective cell type from the plurality of cell types; and determining a plurality of cell composition percentages for the plurality of cell types at least in part by optimizing a piecewise continuous error function between the expression data and the plurality of expression profiles.
- FIG. 1A is a diagram depicting a system for determining a cell composition percentage based on expression data, according to some embodiments of the technology described herein.
- FIG. 1B is an example diagram for determining different cell composition percentages for different cell types and cell subtypes using a non-linear regression model for each respective cell type and cell sub-type, according to some embodiments of the technology described herein.
- FIG. 1C is a t-SNE visualization depicting exemplary cell populations including malignant and microenvironment cells, according to some embodiments of the technology described herein.
- FIG. 1D is a t-SNE visualization depicting exemplary malignant cell populations, according to some embodiments of the technology described herein.
- FIG. 1E is a chart depicting exemplary gene expressions for a variety of cells, according to some embodiments of the technology described herein.
- FIG. 1F is a chart depicting an exemplary correlation between genes and selected cell proportions in a sample mixture of various cell types, according to some embodiments of the technology described herein.
- FIG. 1G is a chart depicting exemplary gene expressions for tumor cell lines, according to some embodiments of the technology described herein.
- FIG. 2A is a flowchart depicting an exemplary non-linear method for determining a cell composition percentage based on expression data, according to some embodiments of the technology described herein.
- FIG. 2B is a flowchart illustrating an example implementation of method 200 for determining a cell composition percentage based on expression data, according to some embodiments of the technology described herein.
- FIG. 2C is a flowchart illustrating an example implementation of act 216 a of method 200 , according to some of the embodiments of the technology described herein.
- FIG. 3A is a diagram depicting use of a machine learning method for determining RNA percentages based on RNA expression data, according to some embodiments of the technology described herein.
- FIG. 3B is a diagram depicting use of a non-linear regression model comprising sub-models for determining RNA percentages based on RNA expression data, according to some embodiments of the technology described herein.
- FIG. 3C is a diagram depicting a method for determining cell composition percentages based on RNA percentages, according to some embodiments of the technology described herein.
- FIG. 3D is a diagram depicting an example method for determining malignancy expression profiles based on cell composition percentages, according to some embodiments of the technology described herein.
- FIG. 4 is a flowchart depicting an exemplary method for training one or more non-linear regression models to determine cell composition percentages based on RNA expression data, according to some embodiments of the technology described herein.
- FIG. 5A-5B are diagrams depicting an exemplary method for training one or more machine learning models including validation and multiple stages of training, according to some embodiments of the technology described herein.
- FIG. 6A is a diagram depicting an exemplary method for training one or more non-linear regression models including generating simulated RNA expression data, according to some embodiments of the technology described herein.
- FIG. 6B is an exemplary diagram for generating artificial mixes of RNA expression data to imitate real tissue, according to some embodiments of the technology described herein.
- FIG. 6C is an exemplary diagram for generating and using artificial mixes to train cell type models, according to some embodiments of the technology described herein.
- FIG. 6D-E are exemplary illustrations for generating specific artificial mixes for training particular cell type/subtype models, according to some embodiments of the technology described herein.
- FIG. 6F is an exemplary diagram illustrating techniques for processing datasets and generating artificial mixes, according to some embodiments of the technology described herein.
- FIG. 7A is a chart comparing simulated RNA expression data to RNA expression data from a biological sample, according to some embodiments of the technology described herein.
- FIG. 7B is a chart depicting exemplary cell composition percentages predicted according to the deconvolution techniques developed by the inventors and corresponding true cell composition percentages, according to some embodiments of the technology described herein.
- FIGS. 7C-7D are charts comparing exemplary prediction accuracy for the deconvolution techniques developed by the inventors, to prediction accuracy for alternative algorithms, according to some embodiments of the technology described herein.
- FIG. 7E is a graph depicting expression of four selected genes in normal tissue, immune cell types, and cancerous tissue, according to some embodiments of the technology described herein.
- FIG. 7F is a chart depicting exemplary prediction specificity for the deconvolution techniques developed by the inventors, according to some embodiments of the technology described herein.
- FIG. 7G is a chart comparing exemplary non-specificity scores for the deconvolution techniques developed by the inventors to non-specificity scores for alternative algorithms, according to some embodiments of the technology described herein.
- FIG. 8 is a flowchart depicting an exemplary linear method for determining cell composition percentages based on RNA expression data, according to some embodiments of the technology described herein.
- FIG. 9A is a diagram depicting exemplary RNA expression profiles and overall RNA expression data, according to some embodiments of the technology described herein.
- FIG. 9B depicts an exemplary piecewise continuous error function, according to some embodiments of the technology described herein.
- FIG. 10 depicts an illustrative implementation of a computer system that may be used in connection with some embodiments of the technology described herein.
- FIG. 11 is a block diagram of an illustrative environment in which one or more embodiments of the technology described herein may be implemented.
- FIGS. 12A-12K are charts and graphs depicting analysis and results from an experiment to establish RNA transcript normalization, and analyze sequencing technical noise as described in connection with Example 1.
- FIGS. 13A-13J are charts and graphs depicting analysis and results from an experiment to deconvolve RNA-seq of multiple normal and cancer tissues as described in connection with Example 2.
- FIGS. 14A-14G are charts and graphs depicting analysis and results from an experiment to deconvolve single cell RNA-seq data and bulk RNA-seq of blood as described in connection with Example 3.
- FIGS. 15A-15I are charts and graphs depicting analysis and results from an experiment to deconvolve several different cancer tissues as described in connection with Example 4.
- determining cell composition percentages e.g., percentages of cells of particular respective types
- a biological sample e.g., such as a sample from a tumor or other diseased tissue
- RNA expression data e.g., data collected by processing the biological sample with a sequencing technique, such as bulk RNA-sequencing.
- determining cell composition percentages for one or more cell types may involve using one or more non-linear regression models to estimate respective cell composition percentages for the cell types.
- the non-linear regression models may be trained using simulated RNA expression data, which may be generated according to the techniques described herein, such as by combining RNA expression data for a variety of malignant and/or microenvironment cell types and/or using any of the sampling, rebalancing, and noising techniques described herein.
- TME tumor microenvironment
- immune and non-immune components of the TME participate in tumor survival, maintenance, growth, and development using cell-to-cell contacts and different molecular signals, such as growth factors and cytokines.
- the inventors have recognized that the TME can mediate tumor survival by controlling the immune system of the host, providing immune surveillance of the tumor. The inventors have therefore appreciated that understanding the quantity and functionality of TME components is essential for cancer research and is important for therapy and understanding its clinical impact.
- RNA-seq bulk RNA-sequencing
- TME cellular composition e.g., cell composition percentages
- cell types may be considered as populations of cells having distinguishable expression profiles.
- CD4+ T cells, CD8+ T cells and NK cells tend to share the expression of a substantial amount of structural and regulatory genes, including metabolic, signaling and surface markers.
- RNA expression data can contain both unique marker genes and genes relevant to the cell lineage.
- the inventors have also recognized that the ratio between marker and lineage-specific gene expression may or may not provide information about cell subtypes (for example the ratio of CD4/CD3D genes may be a marker of CD4+ T cells, but CD3D is not a unique marker for subtypes of helper T cells). Since cells of different types, even if they are closely related, can have significantly different impacts on tumor pathogenesis, the inventors have recognized that it nevertheless may be critical to distinguish cell populations even between closely related cell types.
- Another challenge with cellular deconvolution recognized by the inventors is the difficulty of distinguishing between the number of cells and their state.
- the expression of a gene specific or semi-specific to one cell type may vary depending on the activation state of the cells of that type or may differ between subtypes of that type.
- multiple studies can sequence similar cell subtypes, they may be captured in different biological states.
- the inventors have recognized and appreciated that the variability in biological states can play an important role in developing accurate estimate for cell composition percentages.
- the inventors have recognized and appreciated that the tumor microenvironment may make up only a relatively small fraction of the tumor as a whole.
- the identification of small cell populations from bulk RNA-seq data can be especially challenging because of a reduced signal-to-noise ratio.
- the inventors have recognized that identifying changes in small cell populations (e.g., NK-cells) remains important, as even small cell populations can nevertheless have significant impact on response to treatment.
- RNA enrichment method e.g., total RNA-seq (REF), polyA enriched (REF), exome capture or 3′ scRNA-seq (REF), for example
- scRNA-seq single cell RNA-seq
- a deconvolution method comprising obtaining expression data (e.g., bulk RNA-seq data) for a biological sample from a subject, and determining a cell composition percentages for one or more cell types (e.g., B cells, CD4+ T cells, CD8+ T cells, endothelial cells, fibroblasts, lymphocytes, macrophages, monocytes, NK cells, neutrophils, and T cells).
- expression data e.g., bulk RNA-seq data
- cell types e.g., B cells, CD4+ T cells, CD8+ T cells, endothelial cells, fibroblasts, lymphocytes, macrophages, monocytes, NK cells, neutrophils, and T cells.
- the cell composition percentage may indicate an estimated percentage of cells of a particular respective type in the biological sample.
- determining a cell composition percentage for a particular cell type may comprise obtaining expression data for a set of genes associated with the cell type (e.g., such as one or more marker genes, which may be specific or semi-specific genes for the particular cell type), and processing that expression data with a non-linear regression model to determine the cell composition percentage for the particular cell type.
- this process may be repeated or performed in parallel for each of multiple cell types (which may include subtypes of cell types, as described herein) in order to achieve a deconvolution across the multiple cell types. As described herein at least with respect to FIG. 7 , these techniques present a significant improvement over the prior art.
- machine learning techniques used for determining cell composition percentages may include using multiple non-linear regression models, each trained to determine a cell composition percentage for a particular respective cell type.
- the non-linear regression model may have multiple parameters (e.g., thousands, tens of thousands, hundreds of thousands, at least one million, millions, tens of millions, or hundreds of millions of parameters) and training the non-linear regression model may include estimating values of those parameters, computationally from expression data simulated for training.
- generating the simulated training data may include generating many training sets (e.g., at least 25,000, at least 50,000, at least 100,000, at least 150,000, at least 200,000, at least 500,000, etc.) for each non-linear regression model, for each cell type.
- multiple non-linear regression models may be trained respectively for multiple cells types (e.g., at least 5, at least 10, at least 20, at least 30, at least 40, etc.).
- FIGS. 7C and 7D show that, compared to conventional techniques, the non-linear deconvolution techniques developed by the inventors (e.g., referred to as “Kassandra”) result in more accurate predictions of cell composition percentages for different cell types, even in the presence of cancer cell hyperexpression noise (e.g., as shown in FIG.
- the techniques described herein constitute an improvement to bioinformatics generally and, specifically, to supporting clinical decision making and understanding tumor pathogenesis because the techniques described herein provide for improved methods of determining cell composition percentages (e.g., particularly for cell populations in the tumor microenvironment.)
- the machine learning techniques described herein can successfully identify dependencies and interconnections between genes of phenotypically closely related cell types by using expression data associated with genes that are associated (e.g., specific and/or semi-specific) with the particular subtype as input to a non-linear regression model specifically trained for that subtype, allowing for the accurate detection of cell subtypes even with similar expression patterns ( FIGS. 7A, 7B ).
- the non-linear deconvolution techniques described herein are also more robust than prior algorithms, showing more consistent accuracy across a variety of cell types/subtypes, and providing significantly more accurate results than conventional techniques on realistic, noisy data ( FIGS. 7C, 7D, 13F, 15G ).
- these more accurate results enable improved cancer diagnosis and prognosis, as well as personalized treatment options for the patient.
- the expression data may include expression data associated with particular genes associated with the given cell type.
- the expression data may include expression data associated with genes for a given cell type.
- identifying the genes that are associated with a particular cell type may comprise processing expression data from multiple samples, which may be obtained from multiple databases, and/or with a variety of sequencing techniques, to identify genes that are only or predominantly expressed in certain cell types or subtypes.
- the use of expression data associated with particular genes associated with the particular cell types allows the cellular deconvolution techniques developed by the inventors to leverage domain-specific knowledge relating to which genes are expressed by which cell types, contributing to the success of the techniques described herein.
- a separate non-linear regression model is trained and used to estimate cell composition percentages for each respective cell type and/or subtype being analyzed in a biological sample (e.g., as described herein including at least with respect to FIG. 3A ). This may allow for cell types and/or subtypes in the biological sample to be distinguished more accurately (e.g., as shown in FIGS. 7A-7G ).
- the model architecture may include a tiered structure (e.g., as described herein including at least with respect to FIG.
- the model architecture may include multiple sub-models corresponding to multiple stages, in which the output of one or more previous sub-models (which may comprise, for example, initial predictions of one or more cell composition percentages for one or more cell types) may be used as part of the input for a subsequent sub-model.
- This allows the models to develop more accurate predictions by improving upon their initial predictions (e.g., from a first stage of training and/or using the models) in order to provide a more accurate final predictions (e.g., at a second, third, etc. stage of training and/or using the models).
- a tiered structure may be utilized in which outputs from the first sub-model across multiple models for multiple cell types and/or subtypes may be provided as input to subsequent sub-model(s) for each model.
- first sub-model predictions of cell composition percentages for all cell types may be provided as input to the second sub-models (e.g., for other cell types and/or subtypes.)
- second sub-models e.g., for other cell types and/or subtypes.
- the models described herein have been trained with data representing artificial mixtures of cell types, allowing the training process to take into account the diverse and tissue-specific expression of malignant and microenvironment cells across much larger numbers of samples of diverse composition (e.g., simulating a wide variety of tumor microenvironments) than could be practically possible by physically sampling and analyzing tumor samples.
- This substantially reduces the effort and computational resources associated with training the non-linear regression models for cellular deconvolution.
- the artificial mixes described herein can also be obtained in such a way that they replicate technical noise and capture a wide biological variability, improving the ability of a machine learning model trained using this data to identify biologically meaningful signals in the presence of such noise and variability.
- RNA expression data used to develop these artificial mixes was derived from multiple different samples, across multiple cell populations having a variety of biological states. These artificial mixes improve the ability of the non-linear regression models to effectively estimate cell composition percentages across a variety of cell types in real tumor samples.
- the techniques developed by the inventors also include improved linear techniques for cellular deconvolution.
- one aspect of the linear techniques that contributes to their success is the use of an error function developed by the inventors.
- the error function may be a piecewise, continuous error function. Compared to conventional methods, such as finding a square distance, the piecewise continuous error function accounts for genes that are strongly expressed in tumor cells. This may increase the accuracy for deconvolution of cells in tumor samples.
- the use of such an error function allows the techniques developed by the inventors to more accurately model the error associated with predicted cell composition percentages (e.g., as described herein including with respect to FIGS. 8 and 9A ), providing improved results over conventional techniques.
- FIG. 1A depicts a system 100 for determining cell composition percentages 110 .
- the illustrated system may be implemented in a clinical or laboratory setting.
- the system 100 includes a biological sample 102 , which may be, for example, a tumor biopsy obtained for a subject (e.g., a subject having, suspected of having, or at risk of having cancer).
- a subject may be at risk of having cancer, for example, if the subject has a genetic predisposition (e.g., a known genetic mutation or mutations) to cancer or may have been exposed to cancer causing agents.
- the biological sample 102 may be obtained by performing a biopsy, obtaining a blood sample, a salivary sample, or any other suitable biological sample from the patient.
- the biological sample 102 may have been previously obtained from a subject. Thus any step applied to the sample (e.g., obtaining expression data from the biological sample) may be performed in vitro.
- the biological sample 102 may include diseased tissue (e.g., a tumor), and/or healthy tissue.
- the biological sample may be obtained from a physician, hospital, clinic, or other healthcare provider.
- the origin or preparation methods of the biological sample may include any of the embodiments described with respect to the “Biological Samples” section.
- the subject may include any of the embodiments described with the “Subjects” section.
- the system 100 may further include a sequencing platform 104 , which may produce sequence information 106 .
- the sequencing platform 104 may be a next generation sequencing platform (e.g., IlluminaTM, RocheTM, Ion TorrentTM, etc.), or any high-throughput or massively parallel sequencing platform.
- the sequencing platform 104 may include any suitable sequencing device and/or any sequencing system including one or more devices. In some embodiments, these methods may be automated, in some embodiments, there may be manual intervention.
- the sequence information 106 may be the result of non-next generation sequencing (e.g., Sanger sequencing).
- the sample preparation may be according to manufacturer's protocols.
- the sample preparation may be custom made protocols, or other protocols which are for research, diagnostic, prognostic, and/or clinical purposes.
- the protocols may be experimental.
- the origin or preparation method of the sequence information may be unknown.
- Sequence information 106 can include the sequence data generated by a sequencing protocol (e.g., the series of nucleotides in a nucleic acid molecule identified by next-generation sequencing, sanger sequencing, etc.) as well as information contained therein (e.g., information indicative of source, tissue type, etc.) which may also be considered information that can be inferred or determined from the sequence data. For example, in some embodiments RNA sequence information may be analyzed to determine whether the nucleic acid was primarily polyadenylated or not.
- a sequencing protocol e.g., the series of nucleotides in a nucleic acid molecule identified by next-generation sequencing, sanger sequencing, etc.
- information contained therein e.g., information indicative of source, tissue type, etc.
- RNA sequence information may be analyzed to determine whether the nucleic acid was primarily polyadenylated or not.
- sequence information 106 can include information included in a FASTA file, a description and/or quality scores included in a FASTQ file, an aligned position included in a BAM file, and/or any other suitable information obtained from any suitable file.
- the sequence information 106 may be generated using a nucleic acid from a sample from a subject.
- Reference to a nucleic acid may refer to one or more nucleic acid molecules (e.g., a plurality of nucleic acid molecules).
- the sequence information may be a sequence data indicating a nucleotide sequence of DNA and/or RNA from a previously obtained biological sample of a subject having, suspected of having, or at risk of having a disease.
- the nucleic acid is deoxyribonucleic acid (DNA).
- the nucleic acid is prepared such that the whole genome is present in the nucleic acid.
- the nucleic acid is processed such that only the protein coding regions of the genome remain (e.g., the exome).
- WES whole exome sequencing
- a variety of methods are known in the art to isolate the exome for sequencing, for example, solution based isolation wherein tagged probes are used to hybridize the targeted regions (e.g., exons) which can then be further separated from the other regions (e.g., unbound oligonucleotides). These tagged fragments can then be prepared and sequenced.
- the nucleic acid is ribonucleic acid (RNA).
- sequenced RNA comprises both coding and non-coding transcribed RNA found in a sample.
- total RNA total RNA
- the nucleic acids can be prepared such that the coding RNA (e.g., mRNA) is isolated and used for sequencing. This can be done through any means known in the art, for example by isolating or screening the RNA for polyadenylated sequences. This is sometimes referred to as mRNA-Seq.
- sequence information 106 may include raw DNA or RNA sequence data, DNA exome sequence data (e.g., from whole exome sequencing (WES), DNA genome sequence data (e.g., from whole genome sequencing (WGS)), RNA expression data, gene expression data, bias-corrected gene expression data, or any other suitable type of sequence data comprising data obtained from the sequencing platform 104 and/or comprising data derived from data obtained from sequencing platform 104 .
- the origin or preparation of the sequencing information 106 may include any of the embodiments described with respect to the “Expression Data,” “Obtaining RNA expression data,” “Alignment and annotation,” “Removing non-coding transcripts,” and “Conversion to TPM and gene aggregation” sections.
- the sequence information 106 may be processed using computing device 108 in order to determine cell composition percentages 110 .
- the sequence information 106 may be processed by one or more software programs running on computing device 108 (e.g., as described herein with respect to FIG. 10 ).
- the sequence information 106 may be processed according to the machine-learning based approach of FIGS. 2A-2C , or any other methods described herein for determining cell composition percentages (e.g., such as the non-linear deconvolution methods described at least with respect to FIGS. 2A-2C and 3A-3C and the linear deconvolution methods described at least with respect to FIGS. 8 and 9A -B).
- the computing device 108 may be operated by a user such as a doctor, clinician, researcher, patient, or other individual.
- the user may provide the sequence information 106 as input to the computing device 108 (e.g., by uploading a file), and/or may provide user input specifying processing or other methods to be performed using the sequence information.
- the result may be one or more cell composition percentages 110 .
- each cell composition percentage may represent an estimated percentage of cells of a particular respective type in the biological sample 102 .
- the cell composition percentages are normalized so that the biological sample as a whole represents 100%.
- Cell types may include, for example, B-cells, Plasma B-cells, Non plasma B cells, T cells, CD4+ T-cells, CD8+ T-cells, Treg, T helpers, CD8+ PD1-high, CD8+ PD1-low, NK-cells, monocytes, macrophages, resting tumor associated macrophages (TAM), M1-like or activated macrophages, neutrophils, endothelial cells, and fibroblasts, and/or any other suitable cell types.
- a cell type may comprise one or more subtypes.
- T cells may have subtypes including CD4+ T cells, CD8+ T cells, Tregs, etc.
- the cell composition percentages 110 may include percentages for cell subtypes as well as cell types which are not subtypes of any other cell types. According to some embodiments, the cell composition percentages may include a percentage for an “Other” cell type, which may represent an estimated percentage of cells not accounted for in the other cell composition percentages (e.g., cells of one or more types not explicitly included in the analysis).
- FIG. 1B is an example diagram for determining different cell composition percentages for different cell types and cell subtypes using a non-linear regression model for each respective cell type and cell sub-type, according to some embodiments of the technology described herein.
- a first non-linear regression model, model A 126 may be used to estimate cell composition percentage 128 for cell type A 122 , using sequence information 124 associated with cell type A 122 .
- a second non-linear regression model, model B 136 may be used to estimate cell composition percentage 138 for cell type B 132 , using sequence information 134 associated with cell type B 136 .
- cell type A 122 and cell type B 132 are different cell types.
- cell type A 122 may include B-cells, while cell type B 132 may include T cells.
- cell type A and/or cell type B may be any suitable cell type, as aspects of the technology described herein are not limited in that respect.
- sequence information 124 and sequence information 134 may be obtained for cell type A 122 and cell type B 132 , respectively.
- sequence information may be associated with a set of genes that is specific and/or semi-specific to the cell type.
- sequence information 124 may be associated with a first set of genes that is specific to cell type A 122
- sequence information 134 may be associated with a second set of genes that is specific to cell type B 132 .
- Techniques for identifying genes that are specific and/or semi-specific to a particular cell type and/or subtype may include any of the embodiments described with respect to the “Gene Selection & Specificity” section.
- model A 126 is used to estimate cell composition percentage 128 for cell type A 122
- model B 136 is used to estimate cell composition percentage 138 for cell type B 132
- each of the models may be trained to estimate cell composition percentages for a specific cell type, as described herein including at least with respect to FIG. 4 .
- cell types may include cell subtypes.
- cell subtypes of close origin may share common genes (e.g., with one another and/or with the cell type from which it was differentiated.)
- cell type B 132 includes subtype A 142 and subtype B 162 .
- cell type B 132 may include T cells
- subtype A 142 and subtype B 162 may include subtypes of T cells (e.g., CD4+ and CD8+ T cells).
- a third non-linear regression model, model C 146 may be used to estimate cell composition percentage 148 for subtype A 142 , using sequence information 144 .
- a fourth non-linear regression model, model D 156 may be used to estimate cell composition percentage 158 for subtype B 162 , using sequence information 164 .
- sequence information 144 and sequence information 164 may be obtained for subtype A 142 and subtype B 162 , respectively. In some embodiments, this may include obtaining sequence information associated with a gene set that includes genes specific and/or semi-specific to the subtype. For example, sequence information 144 may be associated with a first set of genes that is specific to subtype A 142 , while sequence information 164 may be associated with a second set of genes that is specific to subtype B 144 . Techniques for identifying genes that are specific and/or semi-specific to a particular cell type and/or subtype may include any of the embodiments described with respect to the “Gene Selection & Specificity” section.
- FIG. 1C is a t-SNE visualization depicting expression data for a plurality of genes for exemplary cell populations including malignant and microenvironment cells.
- the cell types and/or subtypes depicted in the t-SNE plot include macrophages, M1 macrophages, M2 macrophages, B cells, B cells (non-plasma), Plasma B cells, T cells, CD8+ T cells, PD1+ CD8+ T cells, PD1 ⁇ CD8+ T cells, CD4+ T cells, Tregs, T helpers, endothelium cells, monocytes, NK cells, fibroblasts, neutrophils and tumor cells (e.g., cancer cells).
- Malignant cells may comprise tumor cells, or any other cells associated with disease and/or diseased tissue.
- Microenvironment cells may comprise any non-tumor cells, including, for example, immune cells, skin cells, or any other cells not included in the tumor cells.
- RNA-seq samples which may be collected from biological samples via any of the sequencing techniques described herein.
- the RNA-seq datasets may be combined, homogeneously annotated, and bioinformatically recalculated (e.g., expression values are bioinformatically recalculated) to obtain accurate and comparable measurements of transcript expression.
- RNA-seq data was available for 12,450 sorted samples (e.g., sorted by flow cytometry and magnetic-assisted sorting of cells with beads), which could be subdivided into nineteen cell populations of interest. After the removal of low coverage samples and quality checks, the selected samples were distributed between 10 major cell types and 19 cell subpopulations, listed in Table 1, below.
- the quality control techniques may include any of the embodiments described in the “Data collection, analysis, and preprocessing” section, or any other suitable quality control techniques.
- data derived from cells with abnormal physiological states may be identified (e.g., based on the annotations provided with the data) and excluded.
- T cell samples with phorbol myristate acetate/ionomycin activation and/or induced pluripotent stem cell-derived samples were excluded.
- samples with a low isolation purity, sequencing quality parameters, high contamination from other organisms (e.g., organisms other than the primary organism under investigation), and/or low coverage were also eliminated.
- the cell populations may include tumor cells 152 .
- the cancer types may include breast cancer, colorectal cancer, head and neck cancer, kidney cancer, lung cancer, melanoma, pancreatic cancer, prostate cancer, stomach cancer, and/or any other types of cancer.
- some or all of the samples of RNA expression data plotted in FIGS. 1C and 1D may be used as part of selecting specific and/or semi-specific genes for particular cell types/subtypes, as described herein including at least with respect to FIG. 1E .
- some or all of the illustrated samples of RNA expression data may be used as part of generating artificial mixes of RNA expression data, as described herein at least with respect to FIG. 6A .
- RNA expression data may be used.
- similar data sets that include some or all of the cell types represented in Table 1, each represented by a plurality of samples from a plurality of datasets as illustrated in Table 1, can be used.
- FIG. 1E is a heatmap depicting exemplary expressions of genes 170 for cell types 160 .
- the vertical axis represents the cell types 160
- the horizontal axis represents the expression of genes 170 in transcripts per million (TPM).
- TPM transcripts per million
- Each row in the heat map represents a single RNA-seq sample.
- some genes may be considered specific to certain cell types.
- the selected genes 190 may be correlated with the RNA percentage in corresponding sorted cell populations 180 .
- the selected genes 192 may have limited or no expression for tumor cell lines 182 .
- Table 2 specifies, for each of multiple cell types, a set of genes which may be considered specific or semi-specific to that cell type, and/or which may be used for the deconvolution techniques described herein.
- the cellular deconvolution techniques developed by the inventors may involve using only certain gene expression data in order to determine cell composition percentages for a particular cell type.
- only expression data of specific and/or semi-specific genes for the particular cell type may be used, as described herein including at least with respect to FIGS. 2A-2C .
- genes which are highly expressed in malignant cells e.g., cancer cell lines
- may be excluded e.g., specific to tumor cells
- the specific and/or semi-specific genes for a particular cell type e.g., non-malignant cell types
- selecting specific and/or semi-specific genes for a particular cell type may comprise performing any or all of the following techniques: literature analysis, fold change analysis with statistical Kruskal-Wallis test (nonparametric ANOVA analogue), Conover-Iman test (nonparametric pairwise test for multiple comparisons), and/or correlation analysis using the RNA-seq data from FIGS. 1C-1D .
- gene sets may be collected from various sources. In some embodiments, only genes with a known function may be used. Some genes may be similar to the labels used in CYTOF, some may be taken from literature data (which may demonstrate the specificity of certain genes), and/or some genes may be found on existing RNA-seq samples of sorted cells (e.g., after filtering experimental conditions, sequencing quality, and quality by expressions). The search for genes in samples may be carried out in several ways: using differential gene expression, using correlations of gene expression with the proportion of cells in artificial mixes (e.g., as described herein including at least with respect to FIG.
- a gene may be considered “specific” to a particular cell type or subtype if it is only expressed in the particular cell type or cell subtype.
- a gene may be considered “semi-specific” to a particular cell type or subtype when: (1) it is expressed both in the particular cell type or subtype and in one or more other cell types or subtypes; and (2) it is expressed to a greater degree in the particular cell type or subtype than in the other cell type(s) or subtype(s).
- a gene may be considered semi-specific for a particular cell type or subtype if the average expression of the gene in the particular cell type or sub-type is at least a threshold percentage (e.g., 50%, 100%, 200%, 500%, 1000%, etc.) or threshold factor (e.g., a factor of 2, 5, 10, 15, 20, etc.) higher than the average expression of the same gene in the other cell types or sub-types.
- a gene may be considered semi-specific for a particular cell type or subtype if the average expression of the gene in the cell type or subtype is at least ten times higher than the average expression of the gene in the other cell types or subtypes.
- the common genes may be considered semi-specific to the cell types and/or subtypes (e.g., semi-specific to both CD4+ T cells and CD8+ T cells.)
- genes may be selected because their expression is significantly lower or absent in malignant cell (e.g., tumor) lines.
- the specificity criterion can be evaluated when assessed on combined expression data from a plurality of datasets, as described above. In some embodiments, if several types of cells are present in the same dataset, then for each such dataset, a similar specificity analysis may also be carried out inside the datasets to control batch effects.
- TCGA Cancer Genome Atlas
- analysis may be performed to determine how these genes are expressed in TCGA (The Cancer Genome Atlas) for the desired type of tumor. For example, for a given cell type, it may be desirable that the ratios of the average TCGA expression to the average expression lie within a comparable range. In other words, if the average expression of a specific or semi-specific gene (e.g., in a specific or semi-specific set of genes) in TCGA is 70% of the average expression in the samples of the sorted cells, while the other gene expressions of this set are around 5%, then the specific or semi-specific gene is likely expressed by a tumor or other cells, or the cells in the tumor differ greatly in the expression of this gene.
- TCGA Cancer Genome Atlas
- the expression of genes from the same set may be desirable for the expression of genes from the same set to correlate with one another among the TCGA samples for this type of tumor (e.g., the desired type of tumor, above.)
- the mean among the correlations with the other genes from the set may be analyzed.
- the characteristic values of the expression of the considered genes in TCGA LUAD may be low (e.g., less than 10 TPM), so the correlations of these genes with each other may also be low (e.g., due to insufficient sequencing depth). In some cases, there may be especially low gene expressions of NK cells and neutrophils.
- hematopoietic immune cells express CD45 (PTPRC) and HCLS1. Due to their development, immune cells can be divided into lymphocytes and myeloid cells. In turn, lymphocytes can be divided into T, B, NK cells, then CD4+ and CD8+ T cells can be distinguished from among T cells. But among these cells, there are also subtypes that can play an important role both in the development of tumors and in the course of treatment. Therefore, as described herein, it may be desirable for cell composition percentages to be determined for subtypes of certain cells.
- RNA expression data may be difficult, since fewer specific and/or semi-specific genes may be expressed in cell subtypes, and the number of such cells in the tumor microenvironment may be smaller than the combined groups of cells.
- one way to improve the accuracy of determining both cell types and subtypes may be to use information on the expression of genes specific and/or semi-specific for the combined group of cells (e.g., including cell types and cell subtypes that share common genes) in determining cell composition percentages for the cell subtypes.
- genes specific and/or semi-specific for the combined group of cells e.g., including cell types and cell subtypes that share common genes
- Such common genes can be used when determining cell composition percentages of individual cell types and subtypes, for example.
- Another way to use genes common to a group of cell subtypes may be to initially calculate a cell composition percentage for the combined group, then refine that calculation in order to determine cell composition percentages for individual cell types in the group, as described elsewhere herein.
- FIG. 2A is a flowchart depicting a method 200 for determining a cell composition percentage for at least one cell type.
- the method 200 may be carried out on a computing device (e.g., as described herein including at least with respect to FIG. 10 ).
- the computing device may include at least one processor, and at least one non-transitory storage medium storing processor-executable instructions which, when executed, perform the acts of method 200 .
- the method 200 may be carried out, for example, in a system such as system 100 (which may include, for example, a clinical setting or a laboratory setting), by one or more computing devices such as by computing device 108 .
- the method 200 begins with obtaining expression data for a biological sample from a subject.
- obtaining expression data may include obtaining expression data from a biological sample that has been previously obtained from a subject using any suitable techniques.
- obtaining the expression data may include obtaining expression data that has been previously obtained from a biological sample (e.g., obtaining the expression data by accessing a database.)
- the expression data is RNA expression data. Examples of RNA expression data are provided herein.
- the subject may have, be suspected of having, or be at risk of having cancer. As described herein including with respect to FIG.
- the biological sample may comprise a biopsy (e.g., of a tumor or other diseased tissue of the subject), any of the embodiments described herein including with respect to the “Biological Samples” section, or any other suitable type of biological sample.
- the origin or preparation of the expression data may include any of the embodiments described with respect to the “Expression Data” and “Obtaining RNA expression data” sections.
- the expression data may be RNA expression data extracted using any suitable techniques.
- the expression data obtained at act 202 may comprise RNA expression data measured in TPM.
- the expression data may be stored on at least one storage medium and accessed as part of act 202 .
- the expression data may be stored in one or more files or in a database, then read.
- the at least one storage medium storing the RNA expression data may be local to the computing device (e.g., stored on the same at least one non-transitory storage medium), or may be external to the computing device (e.g., stored in a remote database or a cloud storage environment).
- the expression data may be stored on a single storage medium or may be distributed across multiple storage mediums.
- the expression data of act 202 may include first expression data associated with a first set of genes associated with a first cell type (e.g., a cell type of the cell types and/or subtypes being analyzed in the biological sample).
- the first set of genes may comprise genes that are specific and/or semi-specific to the first cell type, as described herein at least with respect to FIG. 1E .
- the set of genes may comprise: ANGPT2, APLN, CDH5, CLEC14A, ECSCR, EMCN, ENG, ESAM, ESM1, FLT1, HHIP, KDR, MMRN1, MMRN2, NOS3, PECAM1, PTPRB, RAWL ROBO4, SELE, TEK, TIE1, and/or VWF.
- the first set of genes may be the same as a set of genes, or a subset of a set of genes, used as part of training a corresponding non-linear regression model for the cell type, as described herein including at least with respect to FIGS. 4-6 .
- determining a first cell composition percentage for the first cell type may comprise processing first expression data associated with a first set of genes for the first cell type with a first non-linear regression model (e.g., of the one or more non-linear regression models) to determine the first cell composition percentage for the first cell type.
- the first expression data may be provided as input to the first non-linear regression model.
- other information may be provided as part of the input to the non-linear regression model.
- a median of the expression data may be included as part of the input to the non-linear regression model.
- any other suitable information may additionally or alternatively be provided as part of the input (e.g., an average of the expression data, a median or average of a subset of the expression data, or any other suitable statistics derived from or otherwise relating to the expression data).
- parts of act 204 may be repeated and/or performed in parallel for each cell type and/or subtype being analyzed.
- a subset of the expression data may be provided as input to each non-linear regression model for each respective cell type and/or subtype.
- the output of the non-linear regression model may comprise information representing an estimated percentage of RNA from the first cell type in the sample.
- the estimate percentage of RNA from the first cell type may be used to calculate a corresponding cell composition percentage for the first cell type.
- the techniques described herein including at least with respect to FIG. 3C may be applied as part of processing the non-linear regression model, such that the output of the non-linear regression model may be an estimated cell composition percentage for the first cell type rather than an estimated percentage of RNA.
- process 200 then proceeds to act 206 for outputting the first cell composition percentage.
- the output(s) of the one or more non-linear regression models may be combined, stored, or otherwise post-processed as part of method 200 .
- the cell composition percentages for each cell type may be stored locally on the computing device used to perform method 200 (e.g., on the non-transitory storage medium).
- the cell composition percentages may be stored in one or more external storage mediums (e.g., such as a remote database or cloud storage environment).
- FIG. 2B is an example implementation of method 200 for determining a cell composition percentage based on expression data.
- implementing method 200 may include any suitable combination of acts included in the example flowchart of FIG. 2B .
- implementing method 200 may include additional or alternative steps that are not shown in FIG. 2B .
- executing method 200 may include every act included in the example flowchart.
- method 200 may include only a subset of the acts included in the example flowchart (e.g., acts 212 and 216 , acts 212 , 214 , 216 , and 218 , acts 212 , 216 , and 220 , etc.).
- the example implementation 220 begins at act 212 , where expression data is obtained for a biological sample from a subject. Obtaining expression data for a biological sample from a subject is described herein above including with respect to act 202 of FIG. 2A .
- act 212 may include obtaining first expression data and second expression data.
- the first expression data may be associated with a first set of genes that is associated with a first cell type, while the second expression data may be associated with a second set of genes that is associated with a second cell type.
- the first expression data may be associated with a first set of genes that is associated with B cells, while the second expression data may be associated with a second set of genes that is associated with T cells.
- the first expression data may be associated with a first set of genes associated with a first cell subtype, while the second expression data may be associated with a second set of genes associated with a second cell subtype.
- the first expression data may be associated with a first set of genes associated with CD4+ cells, while the second expression data may be associated with a second set of genes associated with CD8+ cells.
- Techniques for identifying genes associated with different cell type and/or subtypes are described herein including with respect to the “Gene Selection & Specificity” section.
- the example method 220 proceeds to act 214 , where the expression data is pre-processed.
- the pre-processing may make the expression data suitable to be processed using the one or more non-linear regression models.
- the expression data may be sorted, combined, organized into batches, filtered, or pre-processed with any other suitable techniques.
- techniques for processing the expression data may include any of the embodiments described with respect to the “Alignment and annotation,” “Removing non-coding transcripts,” and “Conversion to TPM and gene aggregation” sections.
- example method 220 proceeds to act 216 , where a plurality of cell composition percentages may be determined for a plurality of cell types using the expression data and one or more non-linear regression models (e.g., at least five, at least ten, at least fifteen, models.)
- each non-linear regression model may be trained according to the techniques described herein including at least with respect to FIGS. 4-6 .
- a separate non-linear regression model may be used to estimate a cell composition percentage for each cell type and/or subtype.
- act 216 may include act 216 a and act 216 b , each of which includes using a separate non-linear regression model trained for determining cell composition percentages for the first and second cell types and/or subtypes, respectively.
- Act 216 a includes determining a first cell composition percentage for the first cell type using the first expression data and a first non-linear regression model.
- Act 216 b includes determining a second cell composition percentage for the second cell type using the second expression data and a second non-linear regression model.
- act 216 may include only one of acts 216 a and 216 b .
- act 216 may include using one or more additional non-linear regression models for determining cell composition percentages for one or more other cell types (e.g., a third cell type or subtype).
- additional non-linear regression models for determining cell composition percentages for one or more other cell types (e.g., a third cell type or subtype).
- An example implementation of act 216 a is described herein including with respect to FIG. 2C .
- example method 220 proceeds to act 218 for outputting the plurality of cell composition percentages.
- the plurality of cell composition percentages may be output through a graphical user interface, saved to memory, transmitted to one or more other computing devices and/or output in any other suitable way.
- techniques may be used to post-process the plurality of cell composition percentages output at act 218 and/or the expression data obtained at act 212 .
- post-processing techniques may include using the cell composition percentages and expression data to determine a malignancy expression profile for the biological sample at act 220 .
- a malignancy expression profile may include information indicative of the expression of malignant cells included in the biological sample. For example, this may include the expression of different genes associated with the malignant cells.
- determining the malignancy expression profile may include (a) estimating the expression profile for TME cells in the biological sample and (b) subtracting the expression of the TME cells from the total expression (e.g., bulk expression data, expression data obtained at act 212 , etc.) of the biological sample.
- total expression e.g., bulk expression data, expression data obtained at act 212 , etc.
- FIG. 2C shows an example implementation of act 216 a for determining, using the first expression data and the first non-linear regression model, a first cell composition percentage for the first cell type.
- the first non-linear regression model may include a first sub-model and/or a second sub-model for processing the first expression data (e.g., as shown in FIG. 3C ).
- the first expression data may include first expression data associated with a first set of genes associated with the first cell type, as well as second expression data associated with a second set of genes associated with the first cell type.
- the example implementation begins at act 232 , for predicting a first value for the estimated percentage of RNA from the first cell type, using a first sub-model.
- the first expression data associated with the first set of genes and/or any other input information may be provided as input to the first sub-model of the non-linear regression model, and the output may be a predicted percentage of RNA from the first cell type.
- the example implementation proceeds to act 234 , for predicting a second value for the estimated percentage of RNA from the first cell type, using a second sub-model.
- the second expression data associated with the second set of genes may be provided as input to the second sub-model of the non-linear expression model in addition to the prediction from the first sub-model and/or any other input information provided at the first sub-model. Additionally or alternatively, the first expression data associated with the first set of genes may be provided as input to the second sub-model.
- predictions from multiple non-linear regression models may be provided as input to the second sub-model of the non-linear regression model for the first cell type.
- the output of the second sub-model of the non-linear regression model may be an estimated percentage of RNA from the first cell type in the sample.
- the output of the second sub-model may comprise the output of the non-linear regression model for the first cell type, in some embodiments.
- the non-linear regression model may comprise more than two sub-models.
- the second sub-model may be repeated any number of times, with the predictions from one or more of the prior sub-models being included as input each time.
- determining the estimated percentage of RNA from the first cell type may include (a) estimating the number of cells of the first type included in the biological sample and (b) estimating the total number of cells included in the biological sample (e.g., using equation 350 .) Estimating the number of cells of the first type may include comparing the estimated percentage of RNA (e.g., R cell of equation 350 ) to an RNA per cell coefficient (e.g., A cell of equation 350 .) Estimating the total number of cells may include estimating the number of cells of each cell type, then summing those values. Techniques for estimating cell composition percentages are described herein including with respect to FIG. 3C .
- FIG. 3A is a diagram depicting an illustrative use of a machine learning method for determining RNA percentages based on RNA expression data.
- RNA expression data from primary tumor samples 302 available on the TCGA database is processed according to the machine learning techniques described herein including at least with respect to FIGS. 2A-2C , in order to arrive at corresponding estimated RNA percentages 306 for T cells, CD4+ T cells, CD8+ T cells.
- the RNA expression data for the tumor samples 302 is obtained from an online database of RNA expression data (e.g., from The Cancer Genome Atlas (TCGA) database, in this example).
- TCGA Cancer Genome Atlas
- the RNA expression data may be obtained from any suitable source, including one or more databases such as TCGA, or directly from a biological sample (e.g., as described herein including at least with respect to FIG. 1A ).
- the RNA expression data may be processed using non-linear regression models 304 .
- the non-linear regression models 304 may be implemented using a gradient boosting technique (e.g., as implemented in XGBoost) as described herein including at least with respect to FIGS. 4-6 .
- non-linear regression models 304 may comprise separate non-linear regression model for each of multiple cell types.
- the non-linear regression models 304 include a non-linear regression model for T cells, a non-linear regression model for CD4+ T cells, and a non-linear regression model for CD8+ T cells. As shown, additional non-linear regression models for one or more additional cell types and/or subtypes may be provided, in some embodiments.
- the input to the non-linear regression models 304 may comprise a select subset of the RNA expression data for each non-linear regression model.
- the input to a non-linear regression model for a particular cell type may comprise RNA expression data for specific and/or semi-specific genes for that cell type.
- the non-linear regression model for T cells may take as input RNA expression data for genes: CAMK4, CBLB, CD2, CD226, CD3D, CD3E, CD3G, CD48, CD5, CD6, CD7, FLT3LG, ITK, KCNA3, KLRB1, LAGS, LAT, LCK, LTA, SIRPG, SIT1, SLA2, TBX21, TCF7, TESPA1, TRAC, TRAF3IP3, TRAT1, TRBC2, TRDC, TRGC1, TRGC2, UBASH3A, ZBED2.
- other information about the RNA expression data e.g., a median of the RNA expression data, or any other suitable statistics
- the output of non-linear regression models 304 may be RNA percentages 306 for respective cell types and/or subtypes.
- the non-linear regression model for T cells may produce as its output a predicted percentage of RNA from T cells in the input RNA expression data.
- the non-linear regression model for CD 4 T cells may produce as its output a predicted percentage of RNA from CD 4 T cells
- the non-linear regression model for CD 8 T cells may produce as its output a predicted percentage of RNA from CD 8 T cells.
- the predicted percentages of RNA may be used to calculate corresponding cell composition percentages for some or all of the cell types and/or subtypes being analyzed.
- the sum of the predictions for the subtypes may or may not be equal to the prediction for the type comprising those subtypes.
- the sum of predictions for CD 4 T cells and CD 8 T cells may exceed the prediction for T cells, or the sum of predictions for CD 4 T cells and CD 8 T cells may be lower than the prediction for T cells.
- the sum of the subtype predictions may be equal to the total type prediction, and/or the subtype predictions may be normalized or adjusted so that their sum is equal to the total type prediction.
- FIG. 3B is a diagram depicting use of non-linear regression models 320 , 322 , 324 comprising first sub-models 326 , 328 , 330 and second sub-models 338 , 340 , 342 for determining RNA percentages based on RNA expression data.
- each example non-linear regression model includes a first sub-model 326 , 328 , 330 , for generating a first value 332 , 334 , 336 for the estimated percentage of RNA from each cell type, and a second sub-model 338 , 340 , 342 for generating a second value 344 , 346 , 348 for the estimated percentage of RNA from each cell type.
- expression data 316 may be obtained from a set of genes associated with cell type B 310 and used as input to the non-linear regression model 322 .
- cell type B 310 may include immune cells and the expression data 316 may include expression data for the genes ADAP2, ADGRE3, ADGRG3, C1QA, C1QC, and C3AR1 (e.g., from the gene set associated with immune cells listed in Table 2.)
- at least some of the expression data 316 e.g., expression data associated with a subset of genes, expression data associated with all the genes, etc.
- the first sub-model may then process the input expression data to determine a first value 334 of the estimated percentage of RNA from cell type B 310 .
- the example non-linear regression model 322 may include a second sub-model 340 to generate a second value 346 of the estimated percentage of RNA from cell type B 310 .
- the second sub-model 340 may use one or more inputs to generate the second value 340 .
- at least some of the expression data 316 may be used as input.
- the expression data may include the same expression data input to the first sub-model 328 (e.g., expression data for the genes ADAP2, ADGRE3, and ADGRG3.) In some embodiments, the expression data may include the same expression data input to the first sub-model, as well as additional expression data (e.g., expression data for the genes ADAP2, ADGRE3, ADGRG3, C1QA, and C3AR1.) In some embodiments, the expression data may include expression data different from the expression data input to the first sub-model (e.g., expression data for the genes C1QA, C1QC, and C3AR1.)
- the second sub-model 340 may take as input estimate percentages of RNA output by the first sub-models 326 , 330 of non-linear regression models 320 , 324 for other cell types 308 , 312 .
- the second sub-model 340 for cell type B 310 takes as input the first value 332 for the estimate percentage of RNA from cell type A 308 and the first value 336 for the estimate percentage of RNA from cell type C 312 .
- This type of input may be informative when trying to determine the percentage of RNA from a cell type that is associated with a same gene or same set of genes as another cell type(s).
- the second sub-model 340 can use the first values 332 , 336 to make such inferences.
- the output of the second sub-model 340 is a second value 346 for the estimated percentage of RNA from cell type B 310 .
- the estimated RNA percentages may be processed to determine cell composition percentages for each of the cell types.
- FIG. 3C is a diagram depicting a method for determining cell composition percentages 370 based on RNA percentages 360 .
- the method of FIG. 3C may be applied to RNA percentages predicted according to the techniques described herein including with respect to FIGS. 2 and 3A , in order to arrive at predictions for cell composition percentages for some or all of the cell types and/or subtypes being analyzed.
- obtaining cell composition percentages based on RNA percentages may comprise applying equation 350 to the RNA percentages for each cell type.
- equation 350 may be applied independently to each RNA percentage (e.g., in sequence), or may be applied to some or all of the RNA percentages together (e.g., in parallel) in some embodiments.
- equation 350 may be applied initially to RNA percentages for cell types which are not subsets of one another.
- equation 350 may subsequently be applied to RNA percentages for cell types that are a subtype of one or more initially used cell types.
- the calculation of cell composition percentages for cell subtypes may be modified based on the initially calculated cell composition percentages. For example, in some embodiments, subsequently calculated cell composition percentages for cell subtypes may be normalized or otherwise adjusted such that they sum to the cell composition percentage for the total cell type (i.e., the initially-calculated cell type of which they are subtypes).
- equation 350 is:
- C c ⁇ e ⁇ l ⁇ l R c ⁇ e ⁇ l ⁇ l A c ⁇ e ⁇ l ⁇ l / ⁇ c ⁇ e ⁇ l ⁇ l ⁇ s ⁇ R c ⁇ e ⁇ l ⁇ l A c ⁇ e ⁇ l ⁇ l ⁇ l
- C cell is the cell composition percentage for the cell type
- R cell is the RNA percentage for the cell type
- a cell is an RNA per cell coefficient.
- the denominator may comprise a sum over all cell types and/or subtypes being analyzed (cells). As such, the expression
- R cell A cell may be initially computed for all cell types and/or subtypes, then used to compute individual C cell values for each cell type and/or subtype.
- an RNA percentage for a cell type may be represented as a fraction or decimal (e.g., for purposes of calculation with equation 350 ).
- an R other expression may be introduced, which may be equal to 1 ⁇ cells R cell .
- equation 350 includes an RNA per cell coefficient A cell , which may represent an RNA concentration per cell.
- the inventors have recognized and appreciated that the abundance of RNA per cell may depend on the cellular size and/or other factors. As such, different cell types may contribute a different amount of RNA to the bulk sample.
- the RNA per cell coefficient can be used to allow the conversion of RNA percentages to corresponding cell composition percentages.
- the RNA per cell coefficient A cell may be determined as part of a model training process (e.g., from simulated or artificial data with known percentages of the different cell types.) In some embodiments, the RNA per cell coefficient A cell may be determined experimentally for some or all cell types.
- RNA per cell coefficients may be obtained by accessing data relating to RNA expression for each cell type (e.g., from available scientific literature, such as PMID: 29130882, PMID: 30726743, or estimated from single cell data, using average or non-linearly transformed UMI count per cell type) and using that data to determine a corresponding RNA per cell coefficient (e.g., by analyzing purity and/or histological TCGA lymphocyte data, for example) for each cell type.
- data relating to RNA expression for each cell type e.g., from available scientific literature, such as PMID: 29130882, PMID: 30726743, or estimated from single cell data, using average or non-linearly transformed UMI count per cell type
- a RNA per cell coefficient e.g., by analyzing purity and/or histological TCGA lymphocyte data, for example
- the RNA per cell coefficients may be tissue specific, and could vary based on the disease being analyzed (e.g., from cancer to cancer). In some embodiments, the RNA per cell coefficient may be tissue agnostic, and may not vary based on a disease being analyzed (e.g., because non-malignant microenvironment cells may be represented by the same or substantially similar cellular phenotypes even across different cancers, tissues, or diseases). In the latter case, data from multiple types of cancers, tissues, diseases, etc. may be combined in order to calculate the RNA per cell coefficients. For example, in some embodiments, more than 10,000 different cancer tissues samples from TCGA were analyzed as part of determining RNA per cell coefficients for cell types.
- determining RNA per cell coefficients may comprise aligning non-malignant cell composition percentages obtained from RNA to cell composition percentages obtained from DNA in order to develop coefficients for RNA per cell type.
- RNA-seq data some embodiments of the technology described herein may be applied to microarray data.
- the expression values may be normalized to lie in a range similar to the values of the transcripts per million (TPM) for RNA-seq (for example, make the sum of the expressions be 1 million) and optionally use a linear scale.
- FIG. 3D is a diagram depicting an example method 380 for determining malignancy expression profiles based on cell composition percentages, according to some embodiments of the technology described herein. This may include obtaining a biological sample (e.g., a biopsy) and determining the expression (e.g., the expression of individual genes) of malignant cells included in the biological sample. In some embodiments, this may include removing the expression of TME cells from the overall expression of the biological sample (e.g., bulk biopsy expression).
- a biological sample e.g., a biopsy
- the expression e.g., the expression of individual genes
- the example method includes three steps.
- the first step 382 includes determining mean expression profiles of different, non-malignant cell types.
- this may include using expression data from sorted cell types.
- this may include obtaining and using RNA-seq data from T cells, B cells, macrophages, fibroblasts, and any other suitable cell type that may be included in a TME.
- the cell types may exclude tumor (e.g., malignant) cells.
- a mean expression profile may include the mean expression of a set of genes for each cell type.
- the example method then proceeds to the second step 384 for predicting the cell composition fractions using cellular deconvolution techniques.
- the cell composition fractions may be indicative of the fraction of each cell type in a biological sample (e.g., a biopsy.) As shown, this may include generating a vector of cell composition fractions.
- cellular deconvolution techniques may include any of the embodiments described herein, including with respect to FIGS. 1-3C .
- the mean expression profiles of different cell types included in the TME (e.g., first step 382 ) and the fraction of each of those cell types in the biological sample (e.g., second step 384 ) may be used to estimate the expression of each cell type in the biological sample.
- the third step 386 may include determining the product of the matrix of expression profiles and the vector of cell fractions. The resulting vector is an estimate expression profile of the TME cells in the biological sample.
- determining the tumor expression profile may include subtracting the TME expression profile from the bulk expression of the biological sample (e.g., the bulk biopsy expression). As shown, this may include subtracting the vector generated for the expression profile of the TME cells from the vector of bulk expression.
- FIG. 4 is a flowchart depicting a method 400 for training one or more non-linear regression models to determine cell composition percentages based on RNA expression data.
- the method 400 may comprise training one or more non-linear regression models (e.g., at least five, at least ten, at least fifteen non-linear regression models) to estimate cell composition percentages for a corresponding one or more cell types in a biological sample.
- a separate non-linear regression model may be trained for each cell type and/or subtype, such that each non-linear regression model is trained to estimate cell composition percentages for a particular cell type in the biological sample.
- the method 400 may be carried out on a computing device (e.g., as described herein including at least with respect to FIG. 10 ).
- the computing device may include at least one processor, and at least one non-transitory storage medium storing processor-executable instructions which, when executed, perform the acts of method 400 .
- the method 400 may begin with obtaining training data comprising simulated RNA expression data.
- the “simulated” RNA expression data may include RNA expression data that is generated partially in silico.
- the simulated RNA expression data may include data that was obtained by sampling reads from multiple expression data sets from purified cell type samples.
- the RNA expression data may comprise expression data measured in TPM.
- the RNA expression data includes first RNA expression data for first genes associated with a first cell type and second RNA expression data for second genes associated with a second cell type.
- the first genes may be, for example, the specific and/or semi-specific genes for the first cell type, while the second genes may be specific and/or semi-specific genes for the second cell type.
- the training data may comprise RNA expression data of genes associated with each cell type and/or subtype being analyzed, and/or other cell types.
- the training data may be generated as part of act 402 .
- the simulated RNA expression data may be generated by combining RNA expression data from malignant cells (e.g., cancer cells) with RNA expression data from microenvironment cells (e.g., immune cells, skin cells, etc.) to produce a plurality of simulated RNA mixtures (which may be referred to herein as “artificial mixtures” or “mixes”) for training.
- RNA mixtures which may be referred to herein as “artificial mixtures” or “mixes”
- at least a thousand, at least ten thousand, at least one hundred thousand, or at least one million mixes may be generated and/or accessed as part of act 402 .
- the training data may be obtained in any suitable manner at act 402 .
- the training data may be stored on at least one storage medium (e.g., in one or more files, or in a database).
- the at least one storage medium storing the training data may be local to the computing device (e.g., stored on the same at least one non-transitory storage medium), or may be external to the computing device (e.g., stored in a remote database or a cloud storage environment).
- the training data may be stored on a single storage medium, or may be distributed across multiple storage mediums.
- act 402 may further comprise pre-processing the training data in any suitable manner.
- the training data may be sorted, combined, organized into batches, filtered, or pre-processed with any other suitable techniques.
- the pre-processing may make the training data suitable to be processed using the one or more non-linear regression models, for example.
- the training data may be split into separate training, validation, and holdout datasets, as described herein including at least with respect to FIG. 5A .
- acts 404 to 408 the method 400 may proceed with training the one or more non-linear regression models using the training data.
- acts 404 to 408 describe training a first model of the non-linear regression models to estimate cell composition percentages for a corresponding first cell type.
- Acts 404 and 406 may be referred to herein as a training step.
- each model of the non-linear regression models may be trained at least in part separately for each cell type (e.g., with corresponding different input data, and different learned parameters, for each non-linear regression model).
- each non-linear regression model of the one or more non-linear regression models may be trained, mutatis mutandis, according to the techniques described herein including with respect to acts 404 to 406 , and/or stored according to act 408 .
- training the first model of the non-linear regression models may proceed with generating an estimated percentage of RNA for the first cell type, using the first model and the first RNA expression data.
- the first RNA expression data may comprise first genes associated with the first cell type (e.g., only specific and/or semi-specific genes for the first cell type).
- the first RNA expression data may be provided as input to the first model.
- other input may additionally or alternatively be provided to the first model. For example, a median, average, or any other suitable information relating the some or all of the RNA expression data may be provided as part of the input to the first model.
- training the first model of the non-linear regression models may proceed with updating parameters using the estimated percentage of RNA from the first cell type.
- the estimated percentage of RNA from the first cell type may be compared to a known value for the percentage of RNA from the first cell type as part of act 406 .
- a loss function may be applied to the estimated value and the known value in order to determine a loss associated with the estimated value.
- the loss may be used to update the parameters of the model. For example, a gradient descent, or any other suitable optimization technique, may be applied in order to update the parameters of the model so as to minimize the loss.
- the first model may process its input using any suitable techniques, including non-linear regression techniques, as described herein.
- the first model may use a gradient boosting machine learning technique.
- the first model may comprise an ensemble of weak prediction models, such as decision trees, or any other suitable prediction models, which may be combined in an iterative fashion using a gradient boosting algorithm.
- a gradient boosting framework such as XGBoost or LightGBM may be used as part of training the first model.
- a random forest model may be used as part of training the first model.
- acts 404 to 406 may be repeated multiple times (e.g., at least one hundred, at least one thousand, at least ten thousand, at least one hundred thousand, or at least one million times). In some embodiments, acts 404 to 406 may be repeated for a set number of iterations, or may be repeated until a threshold is surpassed (e.g., until loss decreases below a threshold value). In some embodiments, the non-linear regression models may be trained in two or more stages, as described herein including at least with respect to FIG. 5A .
- the method 400 may proceed with outputting the trained plurality of non-linear regression models including the first non-linear regression model and the second non-linear regression model.
- outputting the trained plurality of non-linear regression models may comprise: storing one or more of the models in at least one non-transitory computer-readable storage medium (e.g., memory) for subsequent access, providing the model(s) to a recipient (e.g., transmitting data associated with the model(s) to a recipient using any suitable communication network or other means), displaying information associate with the model(s) to a user via a graphical user interface, and/or any other suitable manner of outputting the trained models, as aspects of the technology described herein are not limited in this respect.
- FIG. 5A depicts an exemplary method 500 for training one or more non-linear regression models, according to the techniques developed by the inventors.
- the illustrated techniques may be used in conjunction with any of the other techniques described herein, including at least with respect to FIGS. 2 and 4 .
- the method 500 may begin at act 502 with preparing one or more datasets for training.
- the datasets may be generated (e.g., according to the techniques described herein including at least with respect to FIG. 6A ) and/or accessed (e.g., from one or more databases) as part of act 502 .
- the datasets may comprise a plurality of artificial mixes of RNA expression data, which may comprise RNA expression data from a variety of malignant (e.g., tumor) and/or microenvironment cells.
- the datasets may comprise at least one thousand, at least ten thousand, at least one hundred thousand, or at least one million artificial mixes.
- the datasets may be separated into training datasets and holdout datasets.
- the datasets may be separated into the training and holdout datasets at random in some embodiments, with a set percentage of the datasets to be used for training and holdout, respectively. For instance, in the illustrated example, 80% of the datasets are used as training datasets, while the remaining 20% are retained as holdout datasets.
- the holdout datasets may be used to develop quality metrics (e.g., as described herein including at least with respect to FIG. 7B ). In some embodiments, there may be no holdout datasets, such that all the datasets may be used for training. As shown in the figure at act 502 , the training datasets may be further subdivided into one or more (e.g., ten) folds each containing a respective training and validation set. According to some embodiments, the training datasets may be divided into folds at random. In some embodiments, cross-fold validation may be performed as part of training.
- each non-linear regression model may be trained to estimate, based on input RNA expression data, a corresponding percentage of RNA from a particular cell type.
- the non-linear regression models may be trained in two stages, the first stage corresponding to training a first sub-model of the non-linear regression model, the second stage corresponding to training a second sub-model of the non-linear regression model.
- the first sub-model of each non-linear regression model may be trained to generate an initial prediction for the percentage of RNA from its respective cell type.
- the input may comprise RNA expression data of specific and/or semi-specific genes for the corresponding cell type.
- only the RNA expression data of the specific and/or semi-specific genes for the cell type may be provided as input.
- other information such as a median of the expression data, may be provided.
- the output of the first stage may be initial predictions for the percentages of RNA from each cell type, with each first sub-model of each non-linear regression model providing a prediction for its respective cell type.
- the second sub-model of each non-linear regression model may be trained to generate a second prediction for the percentage of RNA from its respective cell type.
- the input may comprise RNA expression data of specific and/or semi-specific genes for the corresponding cell type, and the predictions from the first stage.
- the RNA expression data used at the second stage may be different from the RNA expression data used at the first stage.
- some or all of the training data may be regenerated (e.g., according to the techniques described herein including with respect to FIGS. 5B and 6 ) for the purposes of training the non-linear regression models in the second stage.
- the training data for the first stage and the second stage may be generated in parallel (e.g., at the same time) but independently, such that the training data for each stage is different.
- the predictions from the first stage may be provided as input at the second stage.
- the initial predictions for all cell types may be provided as input to the second stage. This may allow the second stage to effectively correct the predictions from the first stage, and may increase the consistency and/or accuracy of the final model.
- the output at the second stage may be second predictions for the percentages of RNA from each cell type, with the second sub-model of each non-linear regression model providing a prediction for its respective cell type.
- the second predictions may be the final output of the non-linear regression models (e.g., as described herein including with respect to FIGS. 2 and 4 ).
- additional stages of training e.g., additional sub-models
- may be performed e.g., a third stage, a fourth stage, etc.), with each stage taking as input new training data (e.g., RNA expression data), and the predictions from the previous stage.
- Providing the predictions from the previous stage as part of the input to the next stage may allow a model for a particular cell type to use the information about estimated proportions of other cell types and adapt to them (e.g., by knowing that the total number of T cells equals 10 and number of CD4+ T cells is 8, the number of CD8+ T cells could not exceed 2).
- a multi-stage training procedure as described herein, may allow the model to account for this. This procedure may allow for information from different cell types and subtypes to be used for each individual cell type model.
- FIG. 5B is an exemplary, non-limiting illustration for training a machine learning model, in accordance with some embodiments of the technology described herein.
- the illustrated techniques may be used in conjunction with any of the other techniques described herein, including at least with respect to FIGS. 2 and 4 .
- diagram 530 illustrates the division of the datasets into one or more folds, as described herein including with respect to FIG. 5A .
- the datasets may be randomly split into three folds, with each of the three folds being further divided into a training dataset and a validation dataset.
- datasets may be used to generate artificial mixes, as described herein including with respect to FIG. 6A .
- the folds may then be used to train one or more models for a given set of parameters (e.g., parameters 550 ).
- the parameters may be generated (e.g., at random) based on a set of predetermined ranges shown in Table 3.
- at least some of the (e.g., all) folds may be used to train each cell type model separately.
- validation mixes may be used to evaluate each parameter set and generate associated evaluation data.
- parameters may be updated with each stage of training and/or used as input to subsequent training stages.
- a first fold may be used as input for a first stage of training to generate a first set of parameters.
- a second fold may then be used as input for a second stage of training to generate an updated set of parameters.
- Tables 4 and 5 show example parameters for one or more cell type models after a first stage and a second stage of training, respectively.
- FIG. 6A is a diagram depicting an exemplary method 600 for training one or more non-linear regression models, including generating simulated RNA expression data (e.g., to use as training data, as described herein including at least with respect to FIGS. 4-5 ).
- the simulated RNA expression data may be generated by combining samples of RNA expression data from malignant cells (e.g., cancer cells) and microenvironment cells (e.g., immune cells, stromal cells, etc.), as shown in branches 610 and 620 of the method 600 .
- malignant cells e.g., cancer cells
- microenvironment cells e.g., immune cells, stromal cells, etc.
- An exemplary process for generating artificial mixes of RNA expression data is described herein below with respect to FIG. 6A .
- FIG. 6B is a diagram depicting an example of generating artificial mixes of RNA expression data to imitate real tissue, according to some embodiments of the technology described herein.
- the RNA expression data is derived from one or more sorted cell types/subtypes representing one or more biological states (e.g., positive gene regulation, negative gene regulation, etc.), as shown in branch 630 .
- the one or more cell types/subtypes are mixed in different proportions to generate artificial mixes, as shown in branches 640 and 650 .
- FIG. 6C is an exemplary diagram for generating and using artificial mixes to train cell type models, according to some embodiments of the technology described herein.
- the dataset is divided into folds.
- the resultant datasets are used to create artificial mixes.
- the artificial mixes are used to train and validate each of one or more non-linear regression models that is specific to one or more cell type/subtype.
- the resultant models from each of the folds may be considered together or independently, as described with respect to FIG. 5A .
- FIGS. 6D and 6E are exemplary illustrations for generating specific artificial mixes for training particular cell type/subtype models, according to some embodiments of the technology described herein.
- one or more datasets may be excluded for training a specific cell type/subtype model, as described herein including with respect to Table 6.
- FIG. 6F is an exemplary diagram illustrating techniques for processing datasets and generating artificial mixes, according to some embodiments of the technology described herein.
- act 602 illustrates datasets for a cell type, prior to rebalancing (e.g., resampling large datasets to avoid overtraining models.)
- datasets may be rebalanced 604 and combined into a total set of samples for a specific cell type. Further, as described herein, samples may then be randomly selected in act 608 and averaged in act 612 .
- hyperexpression noise may be added to the expression of the cell type, as illustrated in 614 .
- the samples of RNA expression data may be obtained as described herein including at least with respect to FIGS. 1C-1D .
- a large number of samples of sorted malignant and microenvironment cells may be used to construct the artificial mixes of RNA expression data.
- the number of samples may be on the order of the number of samples included in Table 1.
- the number of samples may be at least 5,000, at least 10,000, at least 15,000, at least 20,000, at least 30,000, at least 50,000, at least 100,000, or any number of suitable samples.
- open source datasets such as Gene Expression Omnibus (GEO) and ArrayExpress may be used.
- GEO Gene Expression Omnibus
- ArrayExpress may be used.
- the datasets used may be selected so as to satisfy the following criteria: only Homo sapiens , standard RNA-seq (without polyA depletion, targeted panel, etc.) with read length higher 31 bp.
- only relevant cell types for the particular disease being analyzed e.g., particular type of tumor
- data for all cell types may instead be used.
- selection of datasets may be based on both biological and bioinformatic parameters. For example, datasets with samples cultivated in conditions close to normal physiological conditions may be used. In some embodiments, datasets with abnormal stimulation were excluded, like datasets of CD4+ T-cells hyper stimulated with phorbol 12-myristate 13-acetate and ionomycin activation or macrophages co-cultured with an excessive number of bacterial cultures. In some embodiments, only those samples having at least 4 million coding read counts were used.
- quality control may be performed on the RNA expression data prior to construction of the artificial mixes (e.g., to exclude strange or unreliable datasets). For example, if some samples of CD4+ T cells show no or very low expression of CD45, CD4 or CD3 genes, they may be excluded. The same may done for other cell types, in some embodiments. For example, samples for some cell types may be excluded if they significantly express genes that are not typical for that type of cell (e.g., if in a sample of T cells, CD19, CD33, MS4A1, etc. were expressed in significant amounts, while in most other T cell samples these expressions were low). In some embodiments, samples of CD4+ T cells may be removed if they express significant amounts of CD8 genes.
- several methods of expression analysis like t-SNE or PCA with different gene sets may be used to visualize the similarities and differences between datasets (e.g., as shown in FIGS. 1C and 1D ). If a particular cell type from one dataset fails to cluster with the same cell type in the other datasets (e.g., in a t-SNE, PCA, or other plot), then the one dataset may be further analyzed as part of quality control, and some or all of the data from that dataset may be excluded.
- RNA expression data may be constructed using samples prepared as described herein above. Artificial mixes may be generated using sample expressions in TPM (transcripts per million) units, such that the gene expressions for an overall sample are formed as a linear combination of the expressions of individual cells from that sample.
- RNA expression data from samples of various cell types may be mixed in predetermined proportions, as described herein below.
- simulated RNA expression data for malignant cells e.g., generated as shown in branch 610
- microenvironment cells e.g., generated as shown in branch 620 .
- samples of each cell type may be rebalanced by datasets (e.g., reducing the weight of datasets with a large number of samples) and subtypes (e.g., changing the proportions of subtypes of a sample).
- datasets e.g., reducing the weight of datasets with a large number of samples
- subtypes e.g., changing the proportions of subtypes of a sample.
- Techniques for rebalancing are described herein including with respect to the “Rebalancing by datasets” and “Rebalancing by subtypes” sections.
- For each cell type multiple samples may then be randomly selected and averaged. Then, for some or all of the cell types being used, the rebalanced/averaged samples may be mixed together in particular proportions (e.g., so as to simulate a real tumor microenvironment).
- branch 610 an exemplary process for generating simulated malignant cell RNA expression data is shown.
- random samples of cancer cells e.g., NSCLC, ccRCC, Mel, HNCK, etc.
- hyperexpression noise may be added to the resulting RNA expression data to account for abnormal expression of genes by malignant cells.
- tumor cells sometimes express genes which are ordinarily absent in the parental cell type.
- the overexpressed genes may interfere with the deconvolution techniques described herein.
- the result of branch 610 may be simulated malignant cell RNA expression data.
- the simulated RNA expression data for the malignant cells (e.g., generated as shown in branch 610 ) and the simulated RNA expression data for the microenvironment cells (e.g., generated as shown in branch 620 ) may be combined into an artificial mix (referred to in FIG. 6A as an “expression mix”).
- the simulated RNA expression data for the malignant cells and the simulated RNA expression data for the microenvironment cells may be mixed together in a random proportion based on a given distribution for cancer cells.
- noise may then be added to the mix to mimic technical noise and noise resulting from biological variability.
- Each type of noise may be specified according to one or more suitable distributions. For example, as shown in FIG.
- the technical noise may be specified by a Poisson distribution, while the noise resulting from biological variability may be specified according to a normal distribution.
- technical noise may have multiple components, which may be specified by other distributions.
- another component of technical noise may be specified by a non-Poisson distribution.
- the artificial mix may be representative of an artificial tumor, including the tumor microenvironment (TME).
- the inventors have recognized and appreciated that, when creating artificial mixes, it may be desirable to use different cells of the same type from different samples. Using a small number of samples for the mixes, or even just one sample for each cell type, would provide poor performance on real tumor samples (e.g., due to the variability of cell states and their expressions, as well as noise due to limited numbers of read counts for different expressions, alignment errors and other causes of technical noise). Therefore, when creating artificial mixtures, the inventors have recognized that is may be desirable to use as many available cell samples as possible.
- RNA-seq samples e.g., at least one hundred, at least five hundred, at least one thousand, at least two thousand, or at least five thousand samples
- a number of datasets of malignant cells e.g., pure cancer cells for various diagnoses, cancer cell lines or sorted from tumors
- Table 7 lists the quantities of samples remaining after quality control for a number of cell types.
- the artificial mixes may be used as training datasets for training one or more non-linear regression models.
- the non-linear regression models may be specific to a cell type/subtype. Accordingly, in some embodiments many (e.g., 150,000) artificial mixes may be generated to train models for each specific cell type model.
- the sets of mixes used for each model may include or exclude specific datasets that allow for differentiation between particular cell types/subtypes, as illustrated in FIGS. 6D and 6E . For example, to train a model for CD4+ T cells, datasets that include unspecified T cells may be excluded to avoid uncertainty about the proportions of CD4+ T cells within the datasets.
- Table 6 specifies the mixes used to train one or more corresponding cell type/subtype models.
- This table specifies, as an example, the samples included in the artificial mixes used to train particular cell type models.
- Mixes set Trained models All Immune cells, available Myeloid cells, samples Lymphocytes, Monocytes, Fibroblasts, Endothelium, Neutrophils, NK cells, T cells, Macrophages, B cells Without T CD4+ T cells, cell samples CD8+ T cells Without T cells, CD8+ T cells CD8+ T PD1 high, CD8 + cells samples T cells PD1 low Without T cells, Tregs, T helpers CD4+ T cells samples Without Macrophages M1, Macrophages Macrophages M2 samples Without Plasma B cells, B cells Non plasma B cells samples Averaging of Samples
- multiple samples for each cell type may be averaged in any suitable manner (e.g., to improve the quality of samples before adding artificial noise). For example, in some embodiments, averaging may be performed in groups of two, such that an averaged sample of 4 million reads may contain information on 8 million reads. In some embodiments, averaging across multiple samples may reduce the noise in the expression caused by technical factors during sequencing.
- num av samples are selected, the expressions of which are averaged (the value of num av is indicated in the parameter table, Table 9).
- any subtype samples may be used at this stage. So, for example, Tregs may be processed along with T cells in some embodiments. Since this approach creates greater subtype diversity for artificial samples but can decrease the biological variability of gene expression within cell type or subtype if too many samples are averaged, the degree of averaging employed may affect the learning outcome. Therefore, the number of samples for averaging may appear as a parameter, which, together with other parameters, may be selected during training (e.g., so as to increase or maximize quality).
- the number of samples may be rebalanced. As described herein below, in one example, the samples may be rebalanced by datasets, then by cell subtypes. Then num av samples may be selected from the rebalanced number of samples.
- the number of samples of sorted cells in datasets may range from one to several hundred (e.g., at least five, at least ten, at least 50, or at least 100 samples).
- each dataset may contain samples of one or two cell types, sorted and sequenced in the same way.
- Cell samples within the same dataset may also have specific conditions, such as a specific set of markers for sorting or a specific disease of patients from whom the cells were taken. Datasets with a large number of samples can lead to overtraining of models for such datasets. To reduce the weight of datasets with a large number of samples, samples of all datasets are resampled in order to rebalance by datasets.
- N d ⁇ a ⁇ t ⁇ a ⁇ s ⁇ e ⁇ t , n ⁇ e ⁇ w N max * ( N d ⁇ a ⁇ t ⁇ a ⁇ s ⁇ e ⁇ t , o ⁇ l ⁇ d N max ) 1 - rebalance ⁇ ⁇ parameter ⁇
- N max is number of samples in the largest dataset (e.g., for the particular cell type) and N dataset,old is the original number of samples in the dataset.
- the rebalance parameter in the equation is a value in the range [0, 1], where 0 means there is no change in the number of samples, and 1 means that for each dataset there will be the same number of samples.
- the rebalancing parameter may be selected during training. Rebalancing by Cell Subtypes
- samples of this type there may also be samples of more specific subtypes.
- the number of available subtype samples may not coincide with those ratios that are specified during the formation of mixes with these subtypes, in some cases. Therefore, when creating mixes for the cell type, samples of its subtypes may be rebalanced.
- CD4+ T cells there may be significantly more CD4+ T cells (and T helpers with Tregs) samples available than CD8+ T cells.
- proportions of CD4+ and CD8+ T cells samples may be changed before the random selection of samples.
- the proportions may be chosen similar to the ratios of the predicted average RNA fractions for the TCGA or PBMC samples for these cell types.
- the predictions may be obtained using one or more linear models trained on mixes with equal cell proportions.
- the subtype rebalancing algorithm may be as follows. To rebalance each subtype for a given type, resample with replacement a number of samples equal to: P subtype *m size/min P +1 Where P subtype is a number reflecting the proportion of a given subtype (e.g., the proportion of this subtype among all subtypes for the given type, which may be represented as the number of samples for the subtype divided by the total number of samples for the type); msize is the maximum number of samples among all the subtypes for the given type, and min P is the minimum number P subtype between all subtypes. According to some embodiments, the rebalancing operation may be performed recursively for all nested subtypes (e.g., subtypes which themselves have subtypes. Microenvironment Cells Proportions Generation
- the resulting samples of different cell types may be mixed with one another in random ratios in order to generate the simulated microenvironment cell RNA expression data.
- a first set of artificial mixes may be generated using random proportions of each cell type:
- R cell is a random number distributed uniformly from 0 to 1 and K cell is the coefficient for the particular cell type.
- the coefficient K cell in the above equations may be chosen so that the most likely ratios of cells mRNA are close to what is observed in TCGA or PBMC samples. These approximate ratios may be calculated from the TCGA or PBMC samples, using models trained without using such ratios. For example, a vector of numbers may be used, reflecting approximate proportions for a given type of tissue. Each number of the vector is multiplied by a random number from 0 to 1. The resulting coefficients are normalized to the sum and used in a linear combination.
- K cell may be selected from Table 7, which specifies, for each of multiple cell types, the most likely proportion of the cell type based on tumor tissue and blood (PBMC).
- the inventors have recognized and appreciated that it may be desirable for the deconvolution algorithm to work in any cell range. For example, the preparation of a cell suspension from a tumor sample may lead to a dramatic increase in the proportion of lymphocytes—and it may be desirable for the algorithm to work on the sequencing data of such a suspension.
- the inventors have recognized and appreciated that the formation of cell ratios by the method described may generate practically no samples where there is a large proportion (e.g., 70-100%) of a certain cell type, such as NK cells. Therefore, in some embodiments, additional mixtures are created in which proportions are generated from the Dirichlet distribution with parameter 1/number_of_types for each dimension. This parameter may be selected along with other parameters for creating mixtures.
- the number of samples in a dataset formed in this way may be controlled by a parameter dirichlet_samples_proportion (Table 9). This parameter may also be selected as a parameter for creating mixtures. Thus, in the final dataset, each cell type may be found in proportions from 0 to 100 percent. However, there most of the characteristic quantities may reflect cell populations that mimic real tumors.
- expressions of artificial tissue may be generated based on expression vectors of each cell type and the randomly selected proportion of RNA of those cells. For example, as described herein, expression vectors are added up with random coefficients that reflect the proportion of RNA of those cells:
- PBMC tumor tissue and blood
- Cell type Solid tumors PBMC B_cells 11 20 Plasma_B_cells 6 3 Non_plasma_B_cells 5 17 T_cells 15 100 CD4_T_cells 7 50 Tregs 4 2 CD8_T_cells 8 50 CD8_T_cells_PD1_low 4 48 CD8_T_cells_PD1_high 4 2 NK_cells 2 16 Monocytes 2 80 Macrophages 40 1 Neutrophils 2 10 Fibroblasts 50 1 Endothelium 36 1 T_helpers 3 48 Macrophages_M1 12 0.5 Macrophages_M2 28 0.5 Noise Generation
- noise e.g., technical noise, uniform noise, or any suitable form of noise
- expression of each gene may contribute noise to the overall tissue expression.
- T i j ⁇ T i +P i j +N prep i +N bio i
- u T i represents the true expression of the gene
- P i j Poisson technical noise
- N prep i represents normally distributed noise derived from sequencing library preparation
- N bio i represents variable biological noise.
- a TPM-based mathematical noise model is provided, which accounts for technical noise (both Poisson and non-Poisson).
- this model of variability may be added to the artificial mixes generated to train the non-linear regression models, as described herein.
- technical non-Poisson noise is assumed to be normally distributed.
- Poisson noise is a type of technical noise which may be associated with the sequencing coverage or number of read counts and may not be normally distributed.
- the resulting dependence of technical noise on coverage and gene expression could be expressed by a formula:
- ⁇ P i ⁇ ⁇ 1 l i ⁇ T ⁇ i ⁇ R
- T i a mean TPM in technical replicates
- R read counts
- ⁇ an estimated proportional coefficient
- this model may correctly represent the gene expression variability as a result of expression levels and coverage, as shown using technical replicates of purified cell populations ( FIG. 12I ).
- the limit of detection of gene expression varied from 1TPM at a coverage of 20 million total reads to 12 TPM at a coverage of 1 million reads per sample Therefore, the ability to assess gene expression may be influenced by the amount of material which is available.
- the noise coefficient (a) can be calculated for Poisson noise ( FIG. 12K ). By calculating this coefficient, the technical noise for each sample and each gene can be inferred according to the deduced formula.
- biological noise which may be associated with different activated states of a cell, can contribute to the overall variance in an RNA-seq sample.
- biological noise may already be present through the use of RNA-seq data derived from cell subsets representing a variation of biological states.
- this overall variance can be assessed, in one example, by plotting data for the same cell types, obtained by different experiments ( FIG. 12J ).
- An example of the dependence of technical noise, both Poisson and non-Poisson, and biological variability on the average sequencing coverage is presented in FIG. 12J .
- the noise increases from 10% to 26% from technical to biological replicates for certain cell types ( FIG. 12J , right).
- the analysis of noise contribution due to single gene expression may be applied to simulate technical and biological noise in artificial mixes.
- noise may be added to total gene expression in two summands:
- T i mix after T i mix before + ⁇ ⁇ T i mix before l i ⁇ ⁇ P + ⁇ ⁇ ⁇ T i mix before ⁇ ⁇ N
- ⁇ P , ⁇ N N(0,1)
- ⁇ is the coefficient of Poisson noise level coefficient
- ⁇ is the coefficient of uniform level non-Poisson noise (Table 9).
- the above-describe approach may be validated by excluding the technical Poisson noise from the technical non-Poisson and biological noise.
- an average variance at about 16% was obtained, which was subsequently used in mixes.
- the noise lost the dependence on the sequencing coverage. This may be expected, since the technical non-Poisson and biological variability do not depend on the measurement method.
- the noise model described herein may be used to add technical (both Poisson and non-Poisson) variation to artificial mixes. This results in artificial mixes which better mimic real tissues. Improved artificial mixes may subsequently be used to train the deconvolution algorithm (e.g., as described herein including with respect to FIGS. 4-6 ) to ensure model stability when encountering real sequencing variability.
- training a non-linear regression model may comprise estimating and/or updating parameters for the model, in some embodiments.
- the parameters for the model may include some parameters, which may be referred to herein as hyperparameters, other than the learned weights for the model (e.g., as described herein at least with respect to FIG. 4 ).
- hyperparameters may be referred to herein as hyperparameters, other than the learned weights for the model (e.g., as described herein at least with respect to FIG. 4 ).
- An exemplary list of such hyperparameters and their values is shown in Table 9.
- values for the hyperparameters may be estimated as the non-linear regression models are trained. For example, some or all of the hyperparameters may be updated based one or more validation sets of the training data (e.g., with each fold of the model training). In some embodiments, the hyperparameters may be estimated based on TCGA data. For example, the results for a particular setting of the hyperparameters may be checked for consistency against TCGA data, such that TCGA model concordance may be achieved.
- the sum of results across the cell subtypes e.g., T cells, B cells, and NK cells
- the sum of results across the cell subtypes is equal (or close) to the overall result for the cell type.
- a parameter search may be performed as part of estimating the hyperparameters. Any suitable parameter search technique may be used, including a random search, a grid search, or a genetic algorithm. In some embodiments, the parameter search may be performed using Bayesian optimization, gradient-based optimization, or evolutionary optimization, for example. In some embodiments, a parameter search may select one or more hyperparameter values from a predetermined range associated with the hyperparameter.
- Tables 8 and 9 list example hyperparameters: number of samples for averaging (Nav), uniform noise level ( ⁇ ), Dirichlet samples proportion (Dp), rebalance parameter (r), hyperexpression fraction (Hf), and maximum hyperexpression level (Mhl).
- a number of artificial mixes “Dp” may be created in which proportion are generated from the Dirichlet distribution.
- the rebalance parameter “r” may be used in an equation to determine a new number of samples in the dataset. As described, “r” is a value in the range [0, 1], where 0 means there is no change in the number of samples, and 1 means that for each dataset there will be the same number of samples. In some embodiments, the rebalancing parameter may be selected during training.
- hyperexpression noise may be added to each of the artificial mixes.
- random values are added to the genes' expression of a selected tumor sample with a small probability for creating each mix. For example, with a probability of “Hf” a random number from a uniform distribution from zero to “Mhl” may be added to the expression of each gene.
- the machine learning models described herein may include tens of thousands, hundreds of thousands, or millions of parameters.
- the non-linear regression models 304 as described herein including at least with respect to FIGS. 2-6 , may include at least ten thousand parameters, at least one hundred thousand parameters, or at least one million parameters.
- processing data with machine learning models like the non-linear regression models 304 requires millions of calculations to be performed, which cannot be done practically in the human mind and without computers.
- the algorithms for training such the machine learning models described herein may require an even greater amount of computational resources, as such models are trained using tens of thousands, hundreds of thousands, or millions of artificial mixes (e.g., as described herein including at least with respect to FIG. 6A ). In one specific example, three million artificial mixes may be generated for training the non-linear regression models across two stages (e.g., as described herein including at least with respect to FIG. 5A ). Neither the training algorithms nor the use of the trained models may be performed without computing resources.
- FIGS. 7A-7G are a variety of results achieved using the techniques developed by the inventors.
- the techniques developed by the inventors substantially outperform conventional techniques for cellular deconvolution.
- the cellular deconvolution techniques developed by the inventors may be referred to as “Kassandra”.
- FIG. 7A is a chart comparing simulated RNA expression data 702 (e.g., a plurality of artificial mixes generated according to the techniques of FIG. 6A ) to RNA expression data 704 from a plurality of biological samples (e.g., tumor).
- the RNA expression data 702 is obtained from five hundred artificial lung cancer samples, developed using the techniques described herein including with respect to FIG. 6A .
- FIG. 7B is a chart depicting exemplary cell composition percentages predicted according to the deconvolution techniques developed by the inventors, and corresponding true cell composition percentages.
- FIGS. 7C and 7D are exemplary charts representing the Pearson correlation for different cell types between predicted and true artificial mix values (e.g., prediction accuracy.)
- the graphs compare exemplary prediction accuracy for the deconvolution techniques developed by the inventors, and the prediction accuracy for alternative algorithms.
- FIG. 7C the prediction accuracy without cancer cell hyperexpression noise is presented.
- FIG. 7D the prediction accuracy with cancer cell hyperexpression is presented.
- random hyperexpression noise may be added to artificial mixes (e.g., to allow the deconvolution techniques developed by the inventors to ignore aberrant expressions from malignant cells in the samples).
- four example gene markers in TCGA data derived from four different cancer types were analyzed: CD14 in bladder cancer, FCRLA in skin cutaneous melanoma, STAP1 in clear cell renal cell carcinoma, and PAD12 in lung squamous cell carcinoma. Each of these markers were found to be overexpressed in the corresponding cancer type. While these markers are not expressed in the corresponding normal tissue, they are found to be expressed in immune cells ( FIG. 7E ).
- the deconvolution techniques developed by the inventors are stable to aberrant high expression present in the data.
- the techniques developed by the inventors produce accurate predictions across cell types, even when hyperexpression noise is present ( FIG. 7D ).
- FIG. 7D indicates that the performance of the alternative algorithms is significantly reduced in the presence of overexpression noise, while the techniques developed by the inventors retained high correlation scores on the validation dataset.
- the alternative algorithms include CIBERSORT, CIBERSORTx, QuanTIseq, FARDEEP, Xcell, ABIS, EPIC, MCP-counter, Scaden, and MuSiC.
- Newman et al. (“Robust enumeration of cell subsets from tissue expression profiles.” Nat. Methods 12, 453-457, (2015)) describes CIBERSORT.
- Newman et al. (“Determining cell type abundance and expression from bulk tissues with digital cytometry.” Nat Biotechnol 37, 773-782 (2019)), describes CIBERSORTx. Finotello et al.
- FIG. 7F is a heatmap representing the Pearson correlation for different cell types between predicted and true artificial mix values (e.g., prediction accuracy) for the deconvolution techniques developed by the inventors. Predicted cell percentages for different cell types are shown for data from sorted samples, derived from holdout datasets. As shown, the deconvolution techniques developed by the inventors achieved high prediction accuracy scores across cell types, including closely related cell types.
- FIG. 7G is a chart comparing exemplary non-specificity scores for the deconvolution techniques developed by the inventors to non-specificity scores for alternative algorithms.
- non-specificity scores for eleven alternative algorithms are shown.
- the values in the chart of FIG. 7G represent percentages of non-specific (false positive) predictions relative to specific (true positive) predictions for different cell types.
- a low non-specificity score indicates a lower percentage of false positives predictions (e.g., indicating a more specific model).
- the detection of signals for each cell type in pure populations was assessed, and B-cells, T-cells, and macrophages were further subdivided, with each subclass clearly distinguished from the others.
- a linear method of cellular deconvolution may be provided.
- An exemplary linear deconvolution technique is described herein below with respect to FIGS. 8 and 9A-9C .
- FIG. 8 is a flowchart depicting an exemplary linear method 800 for determining cell composition percentages based on expression data (e.g., RNA expression data).
- the method 800 may comprise estimating cell composition percentages for one or more cell types in a biological sample, using an expression profile (e.g., an RNA expression, and/or an expression profile as shown in FIG. 9A ) for each cell type.
- an expression profile e.g., an RNA expression, and/or an expression profile as shown in FIG. 9A
- the method 800 may be carried out on a computing device (e.g., as described herein including at least with respect to FIG. 10 ).
- the computing device may include at least one processor, and at least one non-transitory storage medium storing processor-executable instructions which, when executed, perform the acts of method 800 .
- the method 800 may be carried out, for example, in a system such as system 100 (which may include, for example, a clinical setting or a laboratory setting), by one or more computing devices such as by computing device 108 .
- the method 800 may begin with obtaining RNA expression data for a biological sample from a subject.
- act 802 may include accessing RNA expression data that was previously obtained from a biological sample.
- the biological sample may comprise a biopsy (e.g., of a tumor or other diseased tissue of the subject) or any other suitable type of biological sample, and the expression data may be extracted using any suitable techniques.
- the expression obtained at act 802 may comprise RNA expression data measured in TPM.
- the origin or preparation methods of the biological sample may include any of the embodiments described with respect to the “Biological Samples” section.
- the origin or preparation methods of the expression data may include any of the embodiments described with respect to the “Expression Data” and “Obtaining RNA expression data” sections.
- the expression data may be stored on at least one storage medium and accessed as part of act 802 .
- the expression data may be stored in one or more files or in a database, which may be read as part of act 802 .
- the at least one storage medium storing the expression data may be local to the computing device (e.g., stored on the same at least one non-transitory storage medium), or may be external to the computing device (e.g., stored in a remote database or a cloud storage environment).
- the expression data may be stored on a single storage medium or may be distributed across multiple storage mediums.
- act 802 may further comprise pre-processing the expression data in any suitable manner.
- the expression data may be sorted, combined, organized into batches, filtered, or pre-processed with any other suitable techniques.
- the pre-processing may make the expression data suitable to be processed using the linear regression technique described herein with respect to acts 804 - 806 .
- pre-processing the RNA may include any of the embodiments described with respect to the “Alignment and annotation,” “Removing non-coding transcripts,” and “Conversion to TPM and gene aggregation” sections.
- the method 800 may proceed with processing the RNA expression data using a linear regression technique in order to determine one or more corresponding cell composition percentages for the cell types.
- the method 800 may proceed with obtaining a plurality of expression profiles (e.g., as described herein including with respect to FIG. 9A ) for a corresponding plurality of cell types. For example, if CD4+ T cells, NK cells, and CD8+ T cells are being analyzed using method 800 , then an expression profile for CD4+ T cells, an expression profile for NK cells, and an expression profile for CD8+ T cells may be obtained at act 802 .
- Each of the expression profiles (e.g., RNA expression profiles) may comprise respective expression data (e.g., RNA expression data) from one or more genes associated with a respective cell type from the plurality of cell types.
- the genes associated with each respective cell type may be specific and/or semi-specific genes for the cell type.
- the genes associated with each respective cell type may comprise corresponding genes listed in Table 2.
- the corresponding genes may include at least 2 genes, at least 4 genes, at least 6 genes, at least 8 genes, at least 10 genes, at least 12 genes, at least 14 genes, or at least 16 genes included in Table 2.
- the corresponding genes may include fewer than 10,000, fewer than 5,000, fewer than 2,000, fewer than 1,000, fewer than 500, fewer than 250, or fewer than 100 genes.
- the expression profile may be obtained in any suitable manner.
- the expression profile may be stored in one or more files or in a database, which may be read as part of act 804 .
- the at least one storage medium storing the expression profile may be local to the computing device (e.g., stored on the same at least one non-transitory storage medium), or may be external to the computing device (e.g., stored in a remote database or a cloud storage environment).
- the expression profile may be stored on a single storage medium, or may be distributed across multiple storage mediums.
- the method 800 may proceed with determining a plurality of cell composition percentages for the plurality of cell types at least in part by optimizing a piecewise continuous error function (e.g., the example piecewise continuous error function described with respect to FIG. 9A ) between the expression data and the plurality of expression profiles.
- Act 806 may be performed simultaneously or iteratively across the plurality of cell types, and may be repeated (e.g., for a set number of iterations, or until a measurement of error is below a threshold) in some embodiments.
- act 806 may comprise performing a linear regression using the expression data, the plurality of expression profiles, and the piecewise continuous error function. This may include, in some embodiments, optimizing the piecewise continuous error function. In some embodiments, optimizing the piecewise continuous error function is not limited to finding a global maximum or minimum of the piecewise continuous error function, but may also encompass finding a local maximum or minimum within a threshold distance of a global maximum or minimum. For example, act 806 may involve determining a combination (e.g., a weighted sum) of the expression profiles that has a lowest error or an error below a threshold (e.g., with the error measured using the piecewise continuous error function) relative to the expression data.
- a combination e.g., a weighted sum
- act 806 may comprise determining, for each gene associated with the cell type, a corresponding output of a piecewise continuous error function (e.g., such as the error function of FIG. 9C ).
- the piecewise continuous error function may serve to compare an actual measured expression value from real data (e.g., RNA-seq data), against a predicted expression value which may be computed using the gene's expression in the expression profile for the cell type (e.g., as obtained at act 804 ).
- the predicted expression value may be computed as a product of the expression of the gene in the expression profile, and a coefficient ⁇ for the cell type.
- the input to the error function may be the coefficient ⁇ , the expression g of the gene in the input expression data, and the expression p of the gene in the expression profile for the cell type.
- the error function may have coefficients a, b, k, as described herein including with respect to FIG. 9C , which may be updated as part of act 806 .
- act 806 may be performed iteratively or in parallel for some or all of the genes. For example, act 806 may be performed repeatedly across the plurality of cell types until a coefficient ⁇ is found for each cell type such that the piecewise continuous error function is below a threshold or minimized.
- the value of coefficient ⁇ may be determined by finding the coefficient value that minimizes the weighted error sum across all of the genes (e.g., the piecewise error function as described herein including with respect to act 806 and FIG. 9C , summed across all genes).
- the coefficient ⁇ may represent the cell composition percentage for the corresponding cell type (e.g., because ⁇ defines the weight of each expression profile in the weighted sum for the expression data).
- determining the plurality of cell composition percentages for the plurality of cell types may comprise processing the coefficients, such as by normalizing them, in order to obtain corresponding cell composition percentages for each of the plurality of cell types.
- FIG. 9A is a diagram depicting exemplary RNA expression profiles and overall RNA expression data.
- known RNA expression profiles are shown for CD4+ T cells, NK cells, and CD8+ T cells.
- Each RNA expression profile is illustrated as a bar graph, with the horizontal axis representing genes, and the vertical axis representing the expression of those genes. As shown in the figure, each RNA expression profile may be unique for a given cell type.
- the overall observed expression for a biological sample may be considered as a sum of expression profiles for cell types comprising the biological sample.
- each RNA expression profile may be weighted by a coefficient ⁇ , such that the biological sample may be considered as a weighted sum of the RNA expression profiles.
- the sum may further include a term for unknown expression of other cell types. This term may represent expression data that is not accounted for with the weighted sum of RNA expression profiles (e.g., as shown in gray in the observed expression for the biological sample).
- FIG. 9B depicts an exemplary piecewise continuous error function for use with the method of FIG. 8 .
- the error function ⁇ is piecewise, with the coefficients a and b dividing the function into three sections, and coefficient k affecting the shape of the rightmost section of the error function.
- the error may be computed according the illustrated expression.
- a biological sample is obtained from a subject having, suspected of having cancer, or at risk of having cancer.
- the biological sample may be any type of biological sample including, for example, a biological sample of a bodily fluid (e.g., blood, urine or cerebrospinal fluid), one or more cells (e.g., from a scraping or brushing such as a cheek swab or tracheal brushing), a piece of tissue (cheek tissue, muscle tissue, lung tissue, heart tissue, brain tissue, or skin tissue), or some or all of an organ (e.g., brain, lung, liver, bladder, kidney, pancreas, intestines, or muscle), or other types of biological samples (e.g., feces or hair).
- a bodily fluid e.g., blood, urine or cerebrospinal fluid
- cells e.g., from a scraping or brushing such as a cheek swab or tracheal brushing
- a piece of tissue e.g.
- the biological sample is a sample of a tumor from a subject. In some embodiments, the biological sample is a sample of blood from a subject. In some embodiments, the biological sample is a sample of tissue from a subject.
- a sample of a tumor refers to a sample comprising cells from a tumor.
- the sample of the tumor comprises cells from a benign tumor, e.g., non-cancerous cells.
- the sample of the tumor comprises cells from a premalignant tumor, e.g., precancerous cells.
- the sample of the tumor comprises cells from a malignant tumor, e.g., cancerous cells.
- tumors include, but are not limited to, adenomas, fibromas, hemangiomas, lipomas, cervical dysplasia, metaplasia of the lung, leukoplakia, carcinoma, sarcoma, germ cell tumors, and blastoma.
- a sample of blood refers to a sample comprising cells, e.g., cells from a blood sample.
- the sample of blood comprises non-cancerous cells.
- the sample of blood comprises precancerous cells.
- the sample of blood comprises cancerous cells.
- the sample of blood comprises blood cells.
- the sample of blood comprises red blood cells.
- the sample of blood comprises white blood cells.
- the sample of blood comprises platelets. Examples of cancerous blood cells include, but are not limited to, leukemia, lymphoma, and myeloma.
- a sample of blood is collected to obtain the cell-free nucleic acid (e.g., cell-free DNA) in the blood.
- a sample of blood may be a sample of whole blood or a sample of fractionated blood.
- the sample of blood comprises whole blood.
- the sample of blood comprises fractionated blood.
- the sample of blood comprises buffy coat.
- the sample of blood comprises serum.
- the sample of blood comprises plasma.
- the sample of blood comprises a blood clot.
- a sample of a tissue refers to a sample comprising cells from a tissue.
- the sample of the tumor comprises non-cancerous cells from a tissue.
- the sample of the tumor comprises precancerous cells from a tissue.
- tissue including organ tissue or non-organ tissue, including but not limited to, muscle tissue, brain tissue, lung tissue, liver tissue, epithelial tissue, connective tissue, and nervous tissue.
- the tissue may be normal tissue or it may be diseased tissue or it may be tissue suspected of being diseased.
- the tissue may be sectioned tissue or whole intact tissue.
- the tissue may be animal tissue or human tissue.
- Animal tissue includes, but is not limited to, tissues obtained from rodents (e.g., rats or mice), primates (e.g., monkeys), dogs, cats, and farm animals.
- the biological sample may be from any source in the subject's body including, but not limited to, any fluid [such as blood (e.g., whole blood, blood serum, or blood plasma), saliva, tears, synovial fluid, cerebrospinal fluid, pleural fluid, pericardial fluid, ascitic fluid, and/or urine], hair, skin (including portions of the epidermis, dermis, and/or hypodermis), oropharynx, laryngopharynx, esophagus, stomach, bronchus, salivary gland, tongue, oral cavity, nasal cavity, vaginal cavity, anal cavity, bone, bone marrow, brain, thymus, spleen, small intestine, appendix, colon, rectum, anus, liver, biliary tract, pancreas, kidney, ureter, bladder, urethra, uterus, vagina, vulva, ovary, cervix, scrotum, penis, prostate, testicle,
- any of the biological samples described herein may be obtained from the subject using any known technique. See, for example, the following publications on collecting, processing, and storing biological samples, each of which are incorporated herein in its entirety: Biospecimens and biorepositories: from afterthought to science by Vaught et al. (Cancer Epidemiol Biomarkers Prev. 2012 February; 21(2):253-5), and Biological sample collection, processing, storage and information management by Vaught and Henderson (IARC Sci Publ. 2011; (163):23-42).
- the biological sample may be obtained from a surgical procedure (e.g., laparoscopic surgery, microscopically controlled surgery, or endoscopy), bone marrow biopsy, punch biopsy, endoscopic biopsy, or needle biopsy (e.g., a fine-needle aspiration, core needle biopsy, vacuum-assisted biopsy, or image-guided biopsy).
- a surgical procedure e.g., laparoscopic surgery, microscopically controlled surgery, or endoscopy
- bone marrow biopsy e.g., punch biopsy, endoscopic biopsy, or needle biopsy
- needle biopsy e.g., a fine-needle aspiration, core needle biopsy, vacuum-assisted biopsy, or image-guided biopsy.
- one or more than one cell may be obtained from a subject using a scrape or brush method.
- the cell biological sample may be obtained from any area in or from the body of a subject including, for example, from one or more of the following areas: the cervix, esophagus, stomach, bronchus, or oral cavity.
- one or more than one piece of tissue e.g., a tissue biopsy
- the tissue biopsy may comprise one or more than one (e.g., 2, 3, 4, 5, 6, 7, 8, 9, 10, or more than 10) biological samples from one or more tumors or tissues known or suspected of having cancerous cells.
- any of the biological samples from a subject described herein may be stored using any method that preserves stability of the biological sample.
- preserving the stability of the biological sample means inhibiting components (e.g., DNA, RNA, protein, or tissue structure or morphology) of the biological sample from degrading until they are measured so that when measured, the measurements represents the state of the sample at the time of obtaining it from the subject.
- a biological sample is stored in a composition that is able to penetrate the same and protect components (e.g., DNA, RNA, protein, or tissue structure or morphology) of the biological sample from degrading.
- degradation is the transformation of a component from one from to another such that the first form is no longer detected at the same level as before degradation.
- a biological sample e.g., tissue sample
- a “fixed” sample relates to a sample that has been treated with one or more agents or processes in order to prevent or reduce decay or degradation, such as autolysis or putrefaction, of the sample.
- fixative processes include but are not limited to heat fixation, immersion fixation, and perfusion.
- a fixed sample is treated with one or more fixative agents.
- fixative agents include but are not limited to cross-linking agents (e.g., aldehydes, such as formaldehyde, formalin, glutaraldehyde, etc.), precipitating agents (e.g., alcohols, such as ethanol, methanol, acetone, xylene, etc.), mercurials (e.g., B-5, Zenker's fixative, etc.), picrates, and Hepes-glutamic acid buffer-mediated organic solvent protection effect (HOPE) fixatuve.
- cross-linking agents e.g., aldehydes, such as formaldehyde, formalin, glutaraldehyde, etc.
- precipitating agents e.g., alcohols, such as ethanol, methanol, acetone, xylene, etc.
- mercurials e.g., B-5, Zenker's fixative, etc.
- picrates e.g., B-5, Zenker's fixative, etc
- a formalin-fixed biological sample is embedded in a solid substrate, for example paraffin wax.
- the biological sample is a formalin-fixed paraffin-embedded (FFPE) sample.
- FFPE formalin-fixed paraffin-embedded
- the biological sample is stored using cryopreservation.
- cryopreservation include, but are not limited to, step-down freezing, blast freezing, direct plunge freezing, snap freezing, slow freezing using a programmable freezer, and vitrification.
- the biological sample is stored using lyophilization.
- a biological sample is placed into a container that already contains a preservant (e.g., RNALater to preserve RNA) and then frozen (e.g., by snap-freezing), after the collection of the biological sample from the subject.
- a preservant e.g., RNALater to preserve RNA
- such storage in frozen state is done immediately after collection of the biological sample.
- a biological sample may be kept at either room temperature or 4° C. for some time (e.g., up to an hour, up to 8 h, or up to 1 day, or a few days) in a preservant or in a buffer without a preservant, before being frozen.
- Non-limiting examples of preservants include formalin solutions, formaldehyde solutions, RNALater or other equivalent solutions, TriZol or other equivalent solutions, DNA/RNA Shield or equivalent solutions, EDTA (e.g., Buffer AE (10 mM Tris.Cl; 0.5 mM EDTA, pH 9.0)) and other coagulants, and Acids Citrate Dextronse (e.g., for blood specimens).
- special containers may be used for collecting and/or storing a biological sample.
- a vacutainer may be used to store blood.
- a vacutainer may comprise a preservant (e.g., a coagulant, or an anticoagulant).
- a container in which a biological sample is preserved may be contained in a secondary container, for the purpose of better preservation, or for the purpose of avoid contamination.
- any of the biological samples from a subject described herein may be stored under any condition that preserves stability of the biological sample.
- the biological sample is stored at a temperature that preserves stability of the biological sample.
- the sample is stored at room temperature (e.g., 25° C.).
- the sample is stored under refrigeration (e.g., 4° C.).
- the sample is stored under freezing conditions (e.g., ⁇ 20° C.).
- the sample is stored under ultralow temperature conditions (e.g., ⁇ 50° C. to ⁇ 800° C.).
- the sample is stored under liquid nitrogen (e.g., ⁇ 1700° C.).
- a biological sample is stored at ⁇ 60° C. to ⁇ 80° C. (e.g., ⁇ 70° C.) for up to 5 years (e.g., up to 1 month, up to 2 months, up to 3 months, up to 4 months, up to 5 months, up to 6 months, up to 7 months, up to 8 months, up to 9 months, up to 10 months, up to 11 months, up to 1 year, up to 2 years, up to 3 years, up to 4 years, or up to 5 years).
- a biological sample is stored as described by any of the methods described herein for up to 20 years (e.g., up to 5 years, up to 10 years, up to 15 years, or up to 20 years).
- Methods of the present disclosure encompass obtaining one or more biological samples from a subject for analysis.
- one biological sample is collected from a subject for analysis.
- more than one (e.g., 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or more) biological samples are collected from a subject for analysis.
- one biological sample from a subject will be analyzed.
- more than one (e.g., 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or more) biological samples may be analyzed.
- the biological samples may be procured at the same time (e.g., more than one biological sample may be taken in the same procedure), or the biological samples may be taken at different times (e.g., during a different procedure including a procedure 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 days; 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 weeks; 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 months, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 years, or 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 decades after a first procedure).
- a second or subsequent biological sample may be taken or obtained from the same region (e.g., from the same tumor or area of tissue) or a different region (including, e.g., a different tumor).
- a second or subsequent biological sample may be taken or obtained from the subject after one or more treatments and may be taken from the same region or a different region.
- the second or subsequent biological sample may be useful in determining whether the cancer in each biological sample has different characteristics (e.g., in the case of biological samples taken from two physically separate tumors in a patient) or whether the cancer has responded to one or more treatments (e.g., in the case of two or more biological samples from the same tumor or different tumors prior to and subsequent to a treatment).
- each of the at least one biological sample is a bodily fluid sample, a cell sample, or a tissue biopsy sample.
- one or more biological specimens are combined (e.g., placed in the same container for preservation) before further processing.
- a first sample of a first tumor obtained from a subject may be combined with a second sample of a second tumor from the subject, wherein the first and second tumors may or may not be the same tumor.
- a first tumor and a second tumor are similar but not the same (e.g., two tumors in the brain of a subject).
- a first biological sample and a second biological sample from a subject are sample of different types of tumors (e.g., a tumor in muscle tissue and brain tissue).
- a sample from which RNA and/or DNA is extracted is sufficiently large such that at least 2 ⁇ g (e.g., at least 2 ⁇ g, at least 2.5 ⁇ g, at least 3 ⁇ g, at least 3.5 ⁇ g or more) of RNA can be extracted from it.
- the sample from which RNA and/or DNA is extracted can be peripheral blood mononuclear cells (PBMCs).
- PBMCs peripheral blood mononuclear cells
- the sample from which RNA and/or DNA is extracted can be any type of cell suspension.
- a sample from which RNA and/or DNA is extracted is sufficiently large such that at least 1.8 ⁇ g RNA can be extracted from it.
- at least 50 mg e.g., at least 1 mg, at least 2 mg, at least 3 mg, at least 4 mg, at least 5 mg, at least 10 mg, at least 12 mg, at least 15 mg, at least 18 mg, at least 20 mg, at least 22 mg, at least 25 mg, at least 30 mg, at least 35 mg, at least 40 mg, at least 45 mg, or at least 50 mg
- tissue sample is collected from which RNA and/or DNA is extracted.
- tissue sample is collected from which RNA and/or DNA is extracted. In some embodiments, at least 30 mg of tissue sample is collected. In some embodiments, at least 10-50 mg (e.g., 10-50 mg, 10-15 mg, 10-30 mg, 10-40 mg, 20-30 mg, 20-40 mg, 20-50 mg, or 30-50 mg) of tissue sample is collected from which RNA and/or DNA is extracted. In some embodiments, at least 30 mg of tissue sample is collected. In some embodiments, at least 20-30 mg of tissue sample is collected from which RNA and/or DNA is extracted.
- a sample from which RNA and/or DNA is extracted is sufficiently large such that at least 0.2 ⁇ g (e.g., at least 200 ng, at least 300 ng, at least 400 ng, at least 500 ng, at least 600 ng, at least 700 ng, at least 800 ng, at least 900 ng, at least 1 ⁇ g, at least 1.1 ⁇ g, at least 1.2 ⁇ g, at least 1.3 ⁇ g, at least 1.4 ⁇ g, at least 1.5 ⁇ g, at least 1.6 ⁇ g, at least 1.7 ⁇ g, at least 1.8 ⁇ g, at least 1.9 ⁇ g, or at least 2 ⁇ g) of RNA can be extracted from it.
- at least 0.2 ⁇ g e.g., at least 200 ng, at least 300 ng, at least 400 ng, at least 500 ng, at least 600 ng, at least 700 ng, at least 800 ng, at least 900 ng, at least 1 ⁇ g, at least 1.1 ⁇ g,
- a sample from which RNA and/or DNA is extracted is sufficiently large such that at least 0.1 ⁇ g (e.g., at least 100 ng, at least 200 ng, at least 300 ng, at least 400 ng, at least 500 ng, at least 600 ng, at least 700 ng, at least 800 ng, at least 900 ng, at least 1 ⁇ g, at least 1.1 ⁇ g, at least 1.2 ⁇ g, at least 1.3 ⁇ g, at least 1.4 ⁇ g, at least 1.5 ⁇ g, at least 1.6 ⁇ g, at least 1.7 ⁇ g, at least 1.8 ⁇ g, at least 1.9 ⁇ g, or at least 2 ⁇ g) of RNA can be extracted from it.
- at least 0.1 ⁇ g e.g., at least 100 ng, at least 200 ng, at least 300 ng, at least 400 ng, at least 500 ng, at least 600 ng, at least 700 ng, at least 800 ng, at least 900 ng, at least 1
- a subject is a mammal (e.g., a human, a mouse, a cat, a dog, a horse, a hamster, a cow, a pig, or other domesticated animal).
- a subject is a human.
- a subject is an adult human (e.g., of 18 years of age or older).
- a subject is a child (e.g., less than 18 years of age).
- a human subject is one who has or has been diagnosed with at least one form of cancer.
- a cancer from which a subject suffers is a carcinoma, a sarcoma, a myeloma, a leukemia, a lymphoma, or a mixed type of cancer that comprises more than one of a carcinoma, a sarcoma, a myeloma, a leukemia, and a lymphoma.
- Carcinoma refers to a malignant neoplasm of epithelial origin or cancer of the internal or external lining of the body.
- Sarcoma refers to cancer that originates in supportive and connective tissues such as bones, tendons, cartilage, muscle, and fat.
- Myeloma is cancer that originates in the plasma cells of bone marrow.
- Leukemias (“liquid cancers” or “blood cancers”) are cancers of the bone marrow (the site of blood cell production). Lymphomas develop in the glands or nodes of the lymphatic system, a network of vessels, nodes, and organs (specifically the spleen, tonsils, and thymus) that purify bodily fluids and produce infection-fighting white blood cells, or lymphocytes.
- a mixed type of cancer include adenosquamous carcinoma, mixed mesodermal tumor, carcinosarcoma, and teratocarcinoma.
- a subject has a tumor.
- a tumor may be benign or malignant.
- a cancer is any one of the following: skin cancer, lung cancer, breast cancer, prostate cancer, colon cancer, rectal cancer, cervical cancer, and cancer of the uterus.
- a subject is at risk for developing cancer, e.g., because the subject has one or more genetic risk factors, or has been exposed to or is being exposed to one or more carcinogens (e.g., cigarette smoke, or chewing tobacco).
- Expression data (e.g., indicating expression levels) for a plurality of genes may be used for any of the methods or compositions described herein.
- the number of genes which may be examined may be up to and inclusive of all the genes of the subject. In some embodiments, expression levels may be examined for all of the genes of a subject.
- the expression data may include, for each cell type listed in Table 2, expression data for at least 5, at least 10, at least 15, at least 20, at least 25, at least 35, at least 50, at least 75, at least 100 genes selected from the group of genes for that cell type in Table 2.
- any method may be used on a sample from a subject in order to acquire expression data (e.g., indicating expression levels) for the plurality of genes.
- the expression data may be RNA expression data, DNA expression data, or protein expression data.
- DNA expression data refers to a level of DNA in a sample from a subject.
- the level of DNA in a sample from a subject having cancer may be elevated compared to the level of DNA in a sample from a subject not having cancer, e.g., a gene duplication in a cancer patient's sample.
- the level of DNA in a sample from a subject having cancer may be reduced compared to the level of DNA in a sample from a subject not having cancer, e.g., a gene deletion in a cancer patient's sample.
- DNA expression data refers to data for DNA (or gene) expressed in a sample, for example, sequencing data for a gene that is expressed in a patient's sample. Such data may be useful, in some embodiments, to determine whether the patient has one or more mutations associated with a particular cancer.
- RNA expression data may be acquired using any method known in the art including, but not limited to: whole transcriptome sequencing, total RNA sequencing, mRNA sequencing, targeted RNA sequencing, small RNA sequencing, ribosome profiling, RNA exome capture sequencing, and/or deep RNA sequencing.
- DNA expression data may be acquired using any method known in the art including any known method of DNA sequencing. For example, DNA sequencing may be used to identify one or more mutations in the DNA of a subject. Any technique used in the art to sequence DNA may be used with the methods and compositions described herein.
- the DNA may be sequenced through single-molecule real-time sequencing, ion torrent sequencing, pyrosequencing, sequencing by synthesis, sequencing by ligation (SOLiD sequencing), nanopore sequencing, or Sanger sequencing (chain termination sequencing).
- Protein expression data may be acquired using any method known in the art including, but not limited to: N-terminal amino acid analysis, C-terminal amino acid analysis, Edman degradation (including though use of a machine such as a protein sequenator), or mass spectrometry.
- the expression data is acquired through bulk RNA sequencing.
- Bulk RNA sequencing may include obtaining expression levels for one or more genes across RNA extracted from a population of multiple input cells, which population may include multiple different cell types.
- the expression data is acquired through single cell sequencing (e.g., scRNA-seq). Single cell sequencing may include sequencing individual cells.
- the expression data comprises whole exome sequencing (WES) data. In some embodiments, the expression data comprises whole genome sequencing (WGS) data. In some embodiments, the expression data comprises next-generation sequencing (NGS) data. In some embodiments, the expression data comprises microarray data.
- a method to process RNA expression data comprises obtaining RNA expression data for a subject (e.g., a subject who has or has been diagnosed with a cancer).
- obtaining RNA expression data comprises obtaining a biological sample and processing it to perform RNA sequencing using any one of the RNA sequencing methods described herein.
- RNA expression data is obtained from a lab or center that has performed experiments to obtain RNA expression data (e.g., a lab or center that has performed RNA-seq).
- a lab or center is a medical lab or center.
- RNA expression data is obtained by obtaining a computer storage medium (e.g., a data storage drive) on which the data exists.
- RNA expression data is obtained via a secured server (e.g., a SFTP server, or Illumina BaseSpace).
- data is obtained in the form of a text-based filed (e.g., a FASTQ file).
- a file in which sequencing data is stored also contains quality scores of the sequencing data).
- a file in which sequencing data is stored also contains sequence identifier information.
- a method to process RNA expression data comprises aligning and annotating genes in the RNA expression data with known sequences of the human genome to obtain annotated RNA expression data.
- alignment of RNA expression data comprises aligning the data to a known assembled genome for a particular species of subject (e.g., the genome of a human) or to a transcriptome database.
- a sequence alignment software are available and can be used to align data to an assembled genome or a transcriptome database.
- Non-limiting examples of alignment software includes short (unspliced) aligners (e.g., BLAT; BFAST, Bowtie, Burrows-Wheeler Aligner, Short Oligonucleotide Analysis package, or Mosaik), spliced aligners, aligners based on known splice junctions (e.g., Errange, IsoformEx, or Splice Seq), or de novo splice aligner (e.g., ABMapper, BBMap, CRAC, or HiSAT).
- any suitable tool can be used for aligning and annotating data.
- Kallisto github.com/pachterlab/kallisto
- Kallisto is used to align and annotate data.
- a known genome is referred to as a reference genome.
- a reference genome also known as a reference assembly
- a reference genome is a digital nucleic acid sequence database, assembled as a representative example of a species' set of genes.
- human and mouse reference genomes used in any one of the methods described herein are maintained and improved by the Genome Reference Consortium (GRC).
- GPC Genome Reference Consortium
- Non-limiting examples of human reference releases are GRCh38, GRCh37, NCBI Build 36.1, NCBI Build 35, and NCBI Build 34.
- a non-limiting example of transcriptome databased include Transcriptome Shotgun Assembly (TSA).
- annotating RNA expression data comprises identifying the locations of genes and/or coding regions in the data to be processed by comparing it to assembled genomes or transcriptome databases.
- data sources for annotation include GENCODE (www.gencodegenes.org), RefSeq (see e.g., www.ncbi.nlm.nih.gov/refseq/), and Ensembl.
- annotating genes in RNA expression data is based on a GENCODE database (e.g., GENCODE V23 annotation; www.gencodegenes.org).
- a method to process RNA expression data comprises removing non-coding transcripts from annotated RNA expression data. Aligning and annotating RNA expression data allows identification of coding and non-coding reads.
- non-coding reads for transcripts are removed so as to concentrate analysis effort on expression of proteins (e.g., those that may be involved in pathology of cancer).
- removing reads for non-coding transcripts from the data reduces the variance in the data, e.g., in replicates of the same or similar sample (e.g., nucleic acid from the same cells or cell-type).
- non-limiting examples of expression data that is removed include one or more non-coding transcripts (e.g., 10-50, 50-100, 100-1,000, 1,000-2,500, 2,500-5,000 or more non-coding transcripts) that belong to one or more gene groups selected from the list consisting of: pseudogenes, polymorphic pseudogenes, processed pseudogenes, transcribed processed pseudogenes, unitary pseudogenes, unprocessed pseudogenes, transcribed unitary pseudogenes, constant chain immunoglobulin (IG C) pseudogenes, joining chain immunoglobulin (IG J) pseudogenes, variable chain immunoglobulin (IG V) pseudogenes, transcribed unprocessed pseudogenes, translated unprocessed pseudogenes, joining chain T cell receptor (TR J) pseudogenes, variable chain T cell receptor (TR V) pseudogenes, small nuclear RNAs (snRNA), small nucleolar RNAs (snoRNA), microRNAs (miRNA), rib
- information for one or more transcripts for one of more of these types of transcripts can be obtained in a nucleic acid database (e.g., a Gencode database, for example Gencode V23, Genbank database, EMBL database, or other database).
- a fraction e.g., 10%, 20% 30%, 40%, 50%, 60%, 70%, 80%, 90%, 95%, 98%, 99%, or 99.5% or more
- the non-coding transcripts histone-encoding gene, mitochondrial genes, interleukin-encoding genes, collagen-encoding genes, and/or T cell receptor-encoding genes as described herein are removed from aligned and annotated RNA expression data.
- a method to process RNA expression data comprises normalizing RNA expression data per length of transcript (e.g., to transcripts per kilobase million (TPM) format) that is read.
- RNA expression data that is normalized per length of transcript is first aligned and annotated. Conversion of data to TPM allows presentation of expression in the form of concentration, rather than counts, which in turn allows comparison of samples with different total read counts and/or length of reads.
- RNA expression data that is normalized per length of transcript read is then analyzed to obtain gene expression data (expression data per gene).
- gene aggregation comprises combining expression data in reads for transcripts for all isoforms of a gene to obtain expression data for that gene.
- gene aggregation to obtain gene expression data is performed after TPM normalization but before identifying genes that introduce bias. In some embodiments, gene aggregation is performed before conversion of the data to TPM.
- FIG. 10 An illustrative implementation of a computer system 1000 that may be used in connection with any of the embodiments of the technology described herein (e.g., such as the method of FIGS. 2, 4, and 6 ) is shown in FIG. 10 .
- the computer system 1000 includes one or more processors 1010 and one or more articles of manufacture that comprise non-transitory computer-readable storage media (e.g., memory 1020 and one or more non-volatile storage media 1030 ).
- the processor 1010 may control writing data to and reading data from the memory 1020 and the non-volatile storage device 1030 in any suitable manner, as the aspects of the technology described herein are not limited in this respect.
- the processor 1010 may execute one or more processor-executable instructions stored in one or more non-transitory computer-readable storage media (e.g., the memory 1020 ), which may serve as non-transitory computer-readable storage media storing processor-executable instructions for execution by the processor 1010 .
- non-transitory computer-readable storage media e.g., the memory 1020
- processor-executable instructions stored in one or more non-transitory computer-readable storage media (e.g., the memory 1020 ), which may serve as non-transitory computer-readable storage media storing processor-executable instructions for execution by the processor 1010 .
- Computing device 1000 may also include a network input/output (I/O) interface 1040 via which the computing device may communicate with other computing devices (e.g., over a network), and may also include one or more user I/O interfaces 1050 , via which the computing device may provide output to and receive input from a user.
- the user I/O interfaces may include devices such as a keyboard, a mouse, a microphone, a display device (e.g., a monitor or touch screen), speakers, a camera, and/or various other types of I/O devices.
- the techniques described herein may be implemented in the illustrative environment 1100 shown in FIG. 11 .
- one or more biological samples of a subject 1180 may be provided to a laboratory 1170 .
- Laboratory 1170 may process the biological sample(s) to obtain expression data (e.g., DNA, RNA, and/or protein expression data) and/or sequence information and provide it, via network 1110 , to at least one database 1160 that stores information about subject (e.g., patient) 1180 .
- expression data e.g., DNA, RNA, and/or protein expression data
- Network 1110 may be a wide area network (e.g., the Internet), a local area network (e.g., a corporate Intranet), and/or any other suitable type of network. Any of the devices shown in FIG. 11 may connect to the network 1110 using one or more wired links, one or more wireless links, and/or any suitable combination thereof.
- a wide area network e.g., the Internet
- a local area network e.g., a corporate Intranet
- Any of the devices shown in FIG. 11 may connect to the network 1110 using one or more wired links, one or more wireless links, and/or any suitable combination thereof.
- the at least one database 1120 may store expression data and or sequence information for the subject (e.g., patient), medical history data for the subject (e.g., patient), test result data for the subject (e.g., patient), and/or any other suitable information about the subject 1180 .
- Examples of stored test result data for the subject (e.g., patient) include biopsy test results, imaging test results (e.g., MRI results), and blood test results.
- the information stored in at least one database 1120 may be stored in any suitable format and/or using any suitable data structure(s), as aspects of the technology described herein are not limited in this respect.
- the at least one database 1120 may store data in any suitable way (e.g., one or more databases, one or more files).
- the at least one database 1120 may be a single database or multiple databases.
- illustrative environment 1100 includes one or more external databases 1120 , which may store information for patients other than patient 1180 .
- external databases 1160 may store expression data and/or sequence information (of any suitable type) for one or more patients, medical history data for one or more patients, test result data (e.g., imaging results, biopsy results, blood test results) for one or more patients, demographic and/or biographic information for one or more patients, and/or any other suitable type of information.
- external database(s) 1160 may store information available in one or more publicly accessible databases such as TCGA (The Cancer Genome Atlas), one or more databases of clinical trial information, and/or one or more databases maintained by commercial sequencing suppliers.
- the external database(s) 1160 may store such information in any suitable way using any suitable hardware, as aspects of the technology described herein are not limited in this respect.
- the at least one database 1120 and the external database(s) 1160 may be the same database, may be part of the same database system, or may be physically co-located, as aspects of the technology described herein are not limited in this respect.
- server(s) 1140 may access information stored in database(s) 1120 and/or 1160 and use this information to perform processes described herein for determining one or more characteristics of a biological sample (e.g., determining cell composition percentages thereof) and/or of the sequence information.
- server(s) 1140 may include one or multiple computing devices. When server(s) 1140 include multiple computing devices, the device(s) may be physically co-located (e.g., in a single room) or distributed across multi-physical locations. In some embodiments, server(s) 1140 may be part of a cloud computing infrastructure. In some embodiments, one or more server(s) 1140 may be co-located in a facility operated by an entity (e.g., a hospital, research institution) with which doctor 1150 is affiliated. In such embodiments, it may be easier to allow server(s) 1140 to access private medical data for the patient 1180 .
- entity e.g., a hospital, research institution
- the results of the analysis performed by server(s) 640 may be provided to doctor 1150 through a computing device 1130 (which may be a portable computing device, such as a laptop or smartphone, or a fixed computing device such as a desktop computer).
- the results may be provided in a written report, an e-mail, a graphical user interface, and/or any other suitable way.
- the results may be provided to patient 1180 or a caretaker of patient 1180 , a healthcare provider such as a nurse, or a person involved with a clinical trial.
- the results may be part of a graphical user interface (GUI) presented to the doctor 1150 via the computing device 1130 .
- GUI graphical user interface
- the GUI may be presented to the user as part of a webpage displayed by a web browser executing on the computing device 1130 .
- the GUI may be presented to the user using an application program (different from a web-browser) executing on the computing device 1130 .
- the computing device 1130 may be a mobile device (e.g., a smartphone) and the GUI may be presented to the user via an application program (e.g., “an app”) executing on the mobile device.
- an application program e.g., “an app”
- FIG. 12A shows the proportions of Transcripts Per Million (TPM) covering transcripts of different biological types calculated in the different samples of purified B cells sequenced in different laboratories (as an example for a cell type).
- GEO and ArrayExpress IDs of the different datasets of sorted B cells are shown as labels on the X axis.
- the transcript biological type is indicated in the legend (according to GENCODE annotation, version 23).
- variability in total expression belonging to short RNA transcripts strongly skews TPM value distribution of genes of interest due to increased variation resulting from length normalization of short transcripts.
- reads for non-coding transcripts from the data may reduce the variance in the data.
- FIG. 12B shows transcripts distribution by transcript biotype and length, as shown in the legend, of a reference human transcriptome (GENCODE, v23). Proportions of transcript numbers of different length for each biotype in the reference transcriptome are shown (with additional categories of all retained and all removed transcripts in FIG. 12C ).
- a substantial amount of noise was derived from short transcripts of TCR- and BCR-coding genes, annotated in the transcriptome as corresponding to V, D, or J regions. While T- and B-cells produce long transcripts after VDJ recombination, these short transcripts are never synthesized; therefore, different TCR and BCR variants (TCR and BCR repertoires) could not be correctly measured without specific realignment.
- TCR and BCR protein-coding transcripts were excluded from TPM normalization. Excluding non-coding transcripts and transcripts of TCR- and BCR-transcripts may reduce the variance in the data, as shown in FIG. 12B .
- FIG. 12C is a schematic representation of an exemplary process for expression quantification and TPM renormalization.
- TPM expressions of transcripts were calculated by Kallisto (Bray et al. 2016).
- Next non-coding transcripts, transcripts coding for TCR/BCR associated with short V, D or J segments and other transcripts according to their biological properties and quality/evidence information are filtered.
- transcripts are aggregated by genes and normalized on 1 million TPM.
- transcript pseudogene transcript pseudogene
- polymorphic_pseudogene polymorphic_pseudogene
- biological processed_pseudogene types according transcribed_processed_pseudogene, to the unitary_pseudogene
- GENCODE unprocessed_pseudogene (Frankish transcribed_unitary_pseudogene, et al.
- IG_C_pseudogene annotation v23 IG_J_pseudogene, IG_V_pseudogene, transcribed_unprocessed_pseudogene, translated_unprocessed_pseudogene, TR_J_pseudogene, TR_V_pseudogene, snRNA, snoRNA, miRNA, ribozyme, rRNA, Mt_tRNA, Mt_rRNA, scaRNA, retained_intron, sense_intronic, sense_overlapping, nonsense_mediated_decay, non_stop_decay, antisense, lincRNA, macro_lncRNA, processed_transcript, 3prime_overlapping_ncrna, sRNA, misc_RNA, vaultRNA, TEC Transcripts of V, D Transcripts of GENCODE biotypes: and J regions of IG_V_gene, IG_
- FIGS. 12D-12E are violin plots showing the relative standard deviations in expression of 3515 housekeeping genes (Eisenberg and Levanon 2013) for different cell types before (red) and after (blue) transcript filtration and TPM renormalization. Data is grouped based on the library preparation type, using either total RNA-seq ( FIG. 12D ) or polyA RNA-seq ( FIG. 12E ). The indicated P-values are calculated by the two-sided Wilcoxon test. Medians of distributions and rank-biserial correlation coefficients are shown.
- FIG. 12F is a PCA projection of RNA expression of sorted B cells obtained from experiments using either total RNA-seq (green) or polyA RNA-seq (red), before (left) and after (right) proposed transcript filtration and renormalization. As shown, there is a decrease in unwanted batch effects between expression profiles, after the procedure of TPM renormalization described herein. Techniques for TPM normalization are described herein including with respect to the “Conversion to TPM and gene aggregation” section.
- FIG. 12G shows the dependence of relative standard deviation of technical replicates on gene expression levels (TPM). RNA-seq experiments with a total coverage of 1 (pink), 5 (yellow) and 10 (green) million readcounts are presented.
- FIG. 12H shows the dependence of mean standard deviation of gene expression on the total coverage of read counts in RNA-seq.
- the illustrated graph shows samples with sequential additions of noise level: Technical Poisson noise only (blue), all technical noise (yellow), and both technical and biological noise (red).
- FIG. 12H (right) is a violin plot showing the distribution of the same standard deviations of gene expression calculated within samples having different types of noise.
- a component of technical noise may specified by a Poisson distribution
- another component of technical noise may be specified by non-Poisson noise
- biological noise may be specified by a normal distribution.
- FIG. 12I is a plot showing measured Poisson noise coefficients for technical replicates of RNA-seq experiments with different total readcount coverage. Poisson noise is inversely proportional to the square root of the total readcount coverage of RNA-seq data.
- FIG. 12J shows the dependence of mean standard deviation of gene expression on the total coverage of read counts in RNA-seq.
- the illustrated graph shows gene expression with imputed Poisson noise (green) and data for the same samples with all technical noise (yellow).
- FIG. 12J shows the dependence of mean standard deviation of gene expression on the total coverage of read counts in RNA-seq.
- the illustrated graph shows the same data as presented in the left graph after subtraction of the imputed Poisson noise, revealing the non-Poisson addition to the technical noise. This non-Poisson technical noise does not show any dependence to sequencing coverage.
- FIG. 12K shows the dependence of mean standard deviation of gene expression on the total coverage of read counts in RNA-seq.
- the illustrated graph shows gene expression for one cell line across various laboratories and experiments, accounting for both biological and technical noise. Imputed Poisson technical noise calculated for the same samples is represented in green.
- FIG. 12K (right) shows the dependence of mean standard deviation of gene expression on the total coverage of read counts in RNA-seq. The illustrated graph shows gene expression as shown on the left after subtraction of the imputed Poisson noise, revealing the pure biological noise in the samples, which did not depend on sequencing coverage.
- Example 2 Deconvolution of Microenvironment from RNA-Seq of Multiple Normal and Cancer Tissues
- RNA-seq data from multiple normal and cancer tissues.
- the cellular deconvolution techniques developed by the inventors may be referred to as “Kassandra”. Specifically, techniques for selecting specific and/or semi-specific genes for cell types and/or subtypes, generating artificial mixes, training multiple non-linear regression models to determine a plurality of cell composition percentages for a plurality of cell types, using the trained non-linear regression models to determine the cell composition percentages, and other pre-processing and post-processing techniques described herein.
- FIG. 13A is a schematic representation of a validation experiment for deconvolution based on TCGA data. Data on the number of cells obtained by other methods from hematoxylin and eosin (H&E) slides and whole exome sequencing (WES) are used.
- H&E hematoxylin and eosin
- WES whole exome sequencing
- FIG. 13B are violin plots showing distributions of cell composition percentages estimated using the deconvolution techniques (e.g., using trained non-linear regression models) described herein for B-cells, CD4+, CD8+, macrophage, fibroblasts, and endothelium cells in 10,489 tumor biopsies from TCGA. As shown, tumor tissues are split by cancer type in the illustrated example.
- FIG. 13C is a t-SNE plot showing TCGA and GTEX samples calculated based on deconvolved cell percentages.
- FIG. 13D is a graph showing the Pearson correlation between percentages of lymphocytes predicted by the techniques described herein on TCGA RNA-seq data and predicted by machine analysis of histological TCGA data by (Saltz et al. 2018).
- FIG. 13E is a plot showing the correlation of predicted percentages of malignant cells from RNA-seq by the techniques described herein, with tumor purity estimated from WES for 11 TCGA cancer types.
- FIG. 13F is a graph showing Pearson correlations between tumor purity and predicted percentages of malignant cells based on RNA-seq data.
- Tumor data was derived from TCGA.
- the graph shows Pearson correlations for predictions by the techniques described herein, as well as Pearson correlations for predictions by various alternative algorithms. Compared to other algorithms, the non-linear deconvolution techniques developed by the inventors more accurately predicted the percentage of malignant cells, demonstrating an improvement over conventional techniques.
- FIG. 13G is a graph showing Pearson correlations of predicted T cell RNA percentages by the techniques described herein with T cell receptor (CDR3 region of TCR) reads by MiXCR in LUSC TCGA data.
- FIG. 13H is a graph showing Pearson correlations of predicted Plasma B cell RNA percentages by the techniques described herein with B cell receptor (CDR3 region of IgH) reads by MiXCR in LUSC TCGA data.
- FIG. 13I is a graph showing Pearson correlation values for predicted T cell RNA percentages with T cell receptor (CDR3 region of TCR) reads in different cancer types from TCGA data. Predictions by the techniques described herein and predictions by various alternative algorithms are shown. Each data point corresponds to a different cancer type (COAD, KIRC, LUAD, LUSC, READ, SKCM, TNBC).
- FIG. 13J is a graph showing Pearson correlation values for predicted Plasma B cell RNA percentages with B cell receptor (CDR3 region of IgH) reads in different cancer types from TCGA. Predictions by the techniques described herein and predictions by various alternative algorithms are shown. Each data point corresponds to a different cancer type (COAD, KIRC, LUAD, LUSC, READ, SKCM, TNBC).
- FIG. 13B the inventors analyzed the cellular composition of TCGA samples of different tumor types and healthy tissues.
- Five major cell populations were quantified including: B-cells, CD4+ T-cells, CD8+ T-cells, Macrophages, Fibroblasts, and Endothelial cells ( FIG. 13C ). These values agreed with what has been reported. For example, DLBC RNA-seq data showed a strong enrichment for B-cells.
- FIG. 13E-F This analysis supports the ability of the techniques described herein to accurately predict cell population from bulk RNAseq data.
- the proportion of expressed T-cell receptor (TCR) and IgH/L (B cell receptor) sequences in the RNA-seq data correlates with the presence of T or plasma B cells actively producing immunoglobulins.
- the sequences were realigned using MIXCR to measure the abundance and diversity of CDR3 transcripts, associated with different T and plasma B cell clones.
- MIXCR Magnetic Ink Characteristics
- RNA-seq single cell RNA-seq data and bulk RNA-seq of blood data.
- the cellular deconvolution techniques developed by the inventors may be referred to as “Kassandra”. Specifically, techniques for generating artificial mixes, selecting specific and/or semi-specific genes for cell types and/or subtypes, training multiple non-linear regression models to determine a plurality of cell composition percentages for a plurality of cell types, using the trained non-linear regression models to determine the cell composition percentages, and other pre-processing and post-processing techniques described herein.
- FIG. 14A is a schematic representation of a validation experiment for deconvolution using scRNA-seq samples from PBMC.
- the scRNA-seq data was artificially mixed to create a bulk RNA-seq dataset.
- FIG. 14B is a t-SNE plot of cell phenotyping across 9 single-cell PBMC datasets provided by 10 ⁇ Genomics.
- the joined plot was obtained by the Seurat pipeline including SCTransform normalization, batch correction and preceding PCA (Butler et al. 2018; Stuart et al. 2019).
- different cell types and/or subtypes express key cell markers (e.g., specific and/or semi-specific genes) that distinguish them.
- FIG. 14C is a graph showing the correlation between true cell percentages from scRNA-seq of PBMC, and predictions made with the techniques described herein for the bulk RNA-seq mixture.
- FIG. 14D are plots showing correlation of true percentages from scRNA-seq of PBMC and predictions made with the techniques described herein (e.g., using non-linear regression models to determine cell composition percentages) for eight cell subtypes.
- FIG. 14E is a schematic representation of a validation experiment for deconvolution using bulk RNA-seq of PBMC or Whole blood and FACS measurement of the same sample.
- FIGS. 14F-1 and 14F-2 are graphs showing the correlation of predicted cell percentages by the techniques described herein from bulk RNA-seq, and actual cell percentages obtained by flow cytometry measurements for different cell types (CD4+ T cells, CD8+ T cells, NK cells, B cells, monocytes and neutrophils). Datasets that were used for comparison are: GSE107572 (Finotello et al. 2019), GSE115823 (Altman et al. 2019), GSE60424 (Linsley et al. 2014), SDY67 (Zimmermann et al. 2016), GSE127813 (Newman et al. 2019), GSE53655 (Shin et al. 2014), GSE64655 (Hoek et al. 2015). Pearson correlations are shown for all cell types combined.
- RNA-seq which was built from scRNA-seq datasets derived from peripheral blood mononuclear cells (PBMCs)
- FIG. 14A-B A high correlation value was obtained when aligning the true scRNA-seq percentage with the predicted RNA-seq percentage
- FIG. 14D when graphing the correlation for each cell type separately, cell types which are present in a high number have the most significant correlation between true and predicted values.
- FIG. 14E Eight different PBMC samples were analyzed and for each sample the FACS analysis was compared to the predicted cell composition by the techniques described herein. As shown, all analysis presented with a correlation coefficient ranging from 0.900 to 0.984 ( FIGS. 14F-1 and 14F-2 ).
- cellular deconvolution techniques developed by the inventors may be referred to as “Kassandra”. Specifically, techniques for generating artificial mixes, selecting specific and/or semi-specific genes for cell types and/or subtypes, training multiple non-linear regression models to determine a plurality of cell composition percentages for a plurality of cell types, using the trained non-linear regression models to determine the cell composition percentages, and other pre-processing and post-processing techniques described herein.
- FIG. 15A depicts t-SNE plots of cell phenotyping, from left to right, in melanoma (GSE72056)(Tirosh et al. 2016), lung carcinoma (E-MTAB-6149 and E-MTAB-6653)(Lambrechts et al. 2018) and head and neck carcinoma (HNC)(GSE103322)(Puram et al. 2017) single-cell datasets.
- the t-SNE plot for lung carcinoma was obtained by the Seurat pipeline including SCTransform normalization, batch correction and preceding PCA (Butler et al. 2018; Stuart et al. 2019).
- the melanoma and head and neck carcinoma t-SNE plots were obtained by t-SNE transformation of log TPM expression values of cell-type-specific genes.
- FIG. 15B is a schematic representation of a validation experiment using scRNA-seq data derived from cancer tissues. scRNA-seq data was artificially mixed to create a bulk RNA-seq dataset.
- FIGS. 15G and 15H are heatmaps showing mean Pearson correlation values ( FIG. 15G ) and mean MAE (Mean Average Error) scores ( FIG. 15H ) between predicted values from artificial bulk RNA-seq data with true values derived from scRNA-seq data for melanoma, lung carcinoma and HNC.
- results from the techniques described herein are compared with results from alternative algorithms.
- the non-linear regression techniques developed by the inventors are shown to, on average, more accurately predict the cell composition percentages for different cell types with lower mean average error.
- FIG. 15I shows the correlation between predicted cell percentages by the techniques described herein and actual cell percentage obtained by FACS for lymphocytes, fibroblasts and lung adenocarcinoma cell line from dataset GSE121127 (Wang et al. 2018)(top) and CYTOF for bone marrow from dataset GSE120444 (Oetjen et al. 2018)(bottom).
- the Pearson correlation value (r) represents correlation value for all cell types combined.
- FIG. 15A cells from scRNA-seq were annotated manually ( FIG. 15A ) and certain percentages of each cell type were mixed to resemble a bulk-RNA-seq sample (e.g., as described herein above at least with respect to FIG. 6A ). Subsequently these cell percentages were compared with predicted values by the techniques described herein. The ability of the techniques described herein to reconstruct cell composition percentages for each cell type was measured ( FIGS. 15C-F ). The median correlation of cell types reconstruction reached ⁇ 0.97 and was the highest among other methods.
- NCBI Gene Gene ID NCBI Accession Number (s) ACAP1 9744 NM_004288, XM_017005386 ACRBP 84519 NM_032415, NM_001324281 ACTA2 59 NM_001141945, NM_001613, NM_001320855 ADAM28 10863 XM_011544370, XM_006716273, XM_011544369, XM_011544371, XR_949375, XM_005273380, XM_006716274, XM_005273382, XM_017012976, XM_017012974, XR_247120, NM_001304351, NM_014265, NR_130710, XM_017012975, NR_130709, XM_01
- One or more aspects and embodiments of the present disclosure involving the performance of processes or methods may utilize program instructions executable by a device (e.g., a computer, a processor, or other device) to perform, or control performance of, the processes or methods.
- a device e.g., a computer, a processor, or other device
- inventive concepts may be embodied as a computer readable storage medium (or multiple computer readable storage media) (e.g., a computer memory, one or more floppy discs, compact discs, optical discs, magnetic tapes, flash memories, circuit configurations in Field Programmable Gate Arrays or other semiconductor devices, or other tangible computer storage medium) encoded with one or more programs that, when executed on one or more computers or other processors, perform methods that implement one or more of the various embodiments described above.
- the computer readable medium or media can be transportable, such that the program or programs stored thereon can be loaded onto one or more different computers or other processors to implement various ones of the aspects described above.
- computer readable media may be non-transitory media.
- program or “software” are used herein in a generic sense to refer to any type of computer code or set of computer-executable instructions that can be employed to program a computer or other processor to implement various aspects as described above. Additionally, it should be appreciated that according to one aspect, one or more computer programs that when executed perform methods of the present disclosure need not reside on a single computer or processor, but may be distributed in a modular fashion among a number of different computers or processors to implement various aspects of the present disclosure.
- Computer-executable instructions may be in many forms, such as program modules, executed by one or more computers or other devices.
- program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types.
- functionality of the program modules may be combined or distributed as desired in various embodiments.
- data structures may be stored in computer-readable media in any suitable form.
- data structures may be shown to have fields that are related through location in the data structure. Such relationships may likewise be achieved by assigning storage for the fields with locations in a computer-readable medium that convey relationship between the fields.
- any suitable mechanism may be used to establish a relationship between information in fields of a data structure, including through the use of pointers, tags or other mechanisms that establish relationship between data elements.
- the software code can be executed on any suitable processor or collection of processors, whether provided in a single computer or distributed among multiple computers.
- a computer may be embodied in any of a number of forms, such as a rack-mounted computer, a desktop computer, a laptop computer, or a tablet computer, as non-limiting examples. Additionally, a computer may be embedded in a device not generally regarded as a computer but with suitable processing capabilities, including a Personal Digital Assistant (PDA), a smartphone, a tablet, or any other suitable portable or fixed electronic device.
- PDA Personal Digital Assistant
- a computer may have one or more input and output devices. These devices can be used, among other things, to present a user interface. Examples of output devices that can be used to provide a user interface include printers or display screens for visual presentation of output and speakers or other sound generating devices for audible presentation of output. Examples of input devices that can be used for a user interface include keyboards, and pointing devices, such as mice, touch pads, and digitizing tablets. As another example, a computer may receive input information through speech recognition or in other audible formats.
- Such computers may be interconnected by one or more networks in any suitable form, including a local area network or a wide area network, such as an enterprise network, and intelligent network (IN) or the Internet.
- networks may be based on any suitable technology and may operate according to any suitable protocol and may include wireless networks, wired networks or fiber optic networks.
- some aspects may be embodied as one or more methods.
- the acts performed as part of the method may be ordered in any suitable way. Accordingly, embodiments may be constructed in which acts are performed in an order different than illustrated, which may include performing some acts simultaneously, even though shown as sequential acts in illustrative embodiments.
- a reference to “A and/or B”, when used in conjunction with open-ended language such as “comprising” can refer, in one embodiment, to A only (optionally including elements other than B); in another embodiment, to B only (optionally including elements other than A); in yet another embodiment, to both A and B (optionally including other elements); etc.
- the phrase “at least one,” in reference to a list of one or more elements, should be understood to mean at least one element selected from any one or more of the elements in the list of elements, but not necessarily including at least one of each and every element specifically listed within the list of elements and not excluding any combinations of elements in the list of elements.
- This definition also allows that elements may optionally be present other than the elements specifically identified within the list of elements to which the phrase “at least one” refers, whether related or unrelated to those elements specifically identified.
- “at least one of A and B” can refer, in one embodiment, to at least one, optionally including more than one, A, with no B present (and optionally including elements other than B); in another embodiment, to at least one, optionally including more than one, B, with no A present (and optionally including elements other than A); in yet another embodiment, to at least one, optionally including more than one, A, and at least one, optionally including more than one, B (and optionally including other elements); etc.
- the terms “approximately,” “substantially,” and “about” may be used to mean within ⁇ 20% of a target value in some embodiments, within ⁇ 10% of a target value in some embodiments, within ⁇ 5% of a target value in some embodiments, within ⁇ 2% of a target value in some embodiments.
- the terms “approximately,” “substantially,” and “about” may include the target value.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Biotechnology (AREA)
- Genetics & Genomics (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Epidemiology (AREA)
- Bioethics (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Public Health (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Physics (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Algebra (AREA)
- General Engineering & Computer Science (AREA)
- Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US17/200,492 US11315658B2 (en) | 2020-03-12 | 2021-03-12 | Systems and methods for deconvolution of expression data |
US17/707,623 US11587642B2 (en) | 2020-03-12 | 2022-03-29 | Systems and methods for deconvolution of expression data |
US18/082,157 US20230178178A1 (en) | 2020-03-12 | 2022-12-15 | Systems and methods for deconvolution of expression data |
Applications Claiming Priority (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US202062988700P | 2020-03-12 | 2020-03-12 | |
US202063108262P | 2020-10-30 | 2020-10-30 | |
US17/200,492 US11315658B2 (en) | 2020-03-12 | 2021-03-12 | Systems and methods for deconvolution of expression data |
Related Child Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US17/707,623 Continuation US11587642B2 (en) | 2020-03-12 | 2022-03-29 | Systems and methods for deconvolution of expression data |
Publications (2)
Publication Number | Publication Date |
---|---|
US20210287759A1 US20210287759A1 (en) | 2021-09-16 |
US11315658B2 true US11315658B2 (en) | 2022-04-26 |
Family
ID=75396875
Family Applications (3)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US17/200,492 Active US11315658B2 (en) | 2020-03-12 | 2021-03-12 | Systems and methods for deconvolution of expression data |
US17/707,623 Active US11587642B2 (en) | 2020-03-12 | 2022-03-29 | Systems and methods for deconvolution of expression data |
US18/082,157 Pending US20230178178A1 (en) | 2020-03-12 | 2022-12-15 | Systems and methods for deconvolution of expression data |
Family Applications After (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US17/707,623 Active US11587642B2 (en) | 2020-03-12 | 2022-03-29 | Systems and methods for deconvolution of expression data |
US18/082,157 Pending US20230178178A1 (en) | 2020-03-12 | 2022-12-15 | Systems and methods for deconvolution of expression data |
Country Status (7)
Country | Link |
---|---|
US (3) | US11315658B2 (ja) |
EP (2) | EP4118657B1 (ja) |
JP (1) | JP2023518185A (ja) |
AU (1) | AU2021233926A1 (ja) |
CA (1) | CA3175126A1 (ja) |
IL (1) | IL296316A (ja) |
WO (1) | WO2021183917A1 (ja) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20220230707A1 (en) * | 2020-03-12 | 2022-07-21 | Bostongene Corporation | Systems and methods for deconvolution of expression data |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2022232615A1 (en) | 2021-04-29 | 2022-11-03 | Bostongene Corporation | Machine learning techniques for estimating tumor cell expression complex tumor tissue |
CA3220280A1 (en) | 2021-05-18 | 2022-11-24 | Bostongene Corporation | Techniques for single sample expression projection to an expression cohort sequenced with another protocol |
CN114038505B (zh) * | 2021-10-19 | 2024-06-14 | 清华大学 | 一种在线整合多来源单细胞数据的方法和系统 |
WO2023076574A1 (en) | 2021-10-29 | 2023-05-04 | Bostongene Corporation | Tumor microenvironment types in breast cancer |
US20240170096A1 (en) | 2022-11-17 | 2024-05-23 | Bostongene Corporation | Rna-seq immunoprofiling of peripheral blood |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019018684A1 (en) | 2017-07-21 | 2019-01-24 | The Board Of Trustees Of The Leland Stanford Junior University | SYSTEMS AND METHODS FOR ANALYZING MIXED CELL POPULATIONS |
US20190233898A1 (en) | 2015-01-22 | 2019-08-01 | The Board Of Trustees Of The Leland Stanford Junior University | Methods and Systems for Determining Proportions of Distinct Cell Subsets |
US20210151128A1 (en) * | 2018-06-29 | 2021-05-20 | Preferred Networks, Inc. | Learning Method, Mixing Ratio Prediction Method, and Prediction Device |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020142563A1 (en) | 2018-12-31 | 2020-07-09 | Tempus Labs, Inc. | Transcriptome deconvolution of metastatic tissue samples |
US11315658B2 (en) | 2020-03-12 | 2022-04-26 | Bostongene Corporation | Systems and methods for deconvolution of expression data |
-
2021
- 2021-03-12 US US17/200,492 patent/US11315658B2/en active Active
- 2021-03-12 EP EP21716903.6A patent/EP4118657B1/en active Active
- 2021-03-12 WO PCT/US2021/022155 patent/WO2021183917A1/en unknown
- 2021-03-12 CA CA3175126A patent/CA3175126A1/en active Pending
- 2021-03-12 JP JP2022554893A patent/JP2023518185A/ja active Pending
- 2021-03-12 EP EP24161223.3A patent/EP4383262A2/en active Pending
- 2021-03-12 AU AU2021233926A patent/AU2021233926A1/en active Pending
- 2021-03-12 IL IL296316A patent/IL296316A/en unknown
-
2022
- 2022-03-29 US US17/707,623 patent/US11587642B2/en active Active
- 2022-12-15 US US18/082,157 patent/US20230178178A1/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190233898A1 (en) | 2015-01-22 | 2019-08-01 | The Board Of Trustees Of The Leland Stanford Junior University | Methods and Systems for Determining Proportions of Distinct Cell Subsets |
WO2019018684A1 (en) | 2017-07-21 | 2019-01-24 | The Board Of Trustees Of The Leland Stanford Junior University | SYSTEMS AND METHODS FOR ANALYZING MIXED CELL POPULATIONS |
US20210151128A1 (en) * | 2018-06-29 | 2021-05-20 | Preferred Networks, Inc. | Learning Method, Mixing Ratio Prediction Method, and Prediction Device |
Non-Patent Citations (65)
Title |
---|
[No Author Listed], Artyomov Lab Systems Immunology. Maxim N. Artyomov. 2013. https://artyomovlab.wustl.edu/site/index.html [last accessed Jun. 7, 2021]. 3 pages. |
[No Author Listed], Deconvolution of ABsolute Immune Signal. shinyapps.io. 2021. https://giannimonaco.shinyapps.io/ABIS/ [last accessed Jun. 7, 2021]. 1 page. |
[No Author Listed], MCP-counter. CIT. 2021. https://cit.ligue-cancer.net/mcp-counter/ [last accessed Jun. 7, 2021]. 2 pages. |
[No Author Listed], Using EPIC to estimate the proportion of various cell types in bulk samples. EPIC. 2021. http://epic.gfellerlab.org/ [last accessed Jun. 7, 2021]. 1 page. |
Abbas et al. Immune response in silico (IRIS): immune-specific genes identified from a compendium of microarray expression data Genes and Immunity vol. 6 pp. 319-331 (Year: 2005). * |
Altman et al., Transcriptome networks identify mechanisms of viral and nonviral asthma exacerbations in children. Nature immunology. May 2019;20(5):637-51. |
Aran et al., Systematic pan-cancer analysis of tumour purity. Nature communications. Dec. 4, 2015;6(1):1-11. |
Aran et al., xCell. UCSF Institute for Computational Health Sciences. 2017. https://xcell.ucsf.edu/ [last accessed Jun. 7, 2021]. 2 pages. |
Aran et al., xCell: digitally portraying the tissue cellular heterogeneity landscape. Genome biology. Dec. 2017;18(1):1-14. |
Becht et al., Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression. Genome biology. Dec. 2016;17(1):1-20. |
Ben-Moshe et al., mRNA-seq whole transcriptome profiling of fresh frozen versus archived fixed tissues. BMC genomics. Dec. 2018;19(1):11 pages. |
Bray et al., Near-optimal probabilistic RNA-seq quantification. Nature biotechnology. May 2016;34(5):525-7. |
Butler et al., Integrating single-cell transcriptomic data across different conditions, technologies, and species. Nature biotechnology. May 2018;36(5):411-20. |
Chen et al., Profiling tumor infiltrating immune cells with CIBERSORT. Cancer systems biology. 2018:243-259. |
Cieslik et al., The use of exome capture RNA-seq for highly degraded RNA with application to clinical cancer sequencing. Genome research. Sep. 1, 2015;25(9):1372-81. |
Eisenberg et al., Human housekeeping genes, revisited. TRENDS in Genetics. Oct. 1, 2013;29(10):569-74. |
Finotello et al., Molecular and pharmacological modulators of the tumor immune contexture revealed by deconvolution of RNA-seq data. Genome medicine. Dec. 2019;11(1):1-20. |
Finotello et al., quanTIseq documentation. quanTIseq. Feb. 25, 2019. https://icbi.i-med.ac.at/software/quantiseq/doc/ [last accessed Jun. 7, 2021]. 9 pages. |
Frankish et al., GENCODE reference annotation for the human and mouse genomes. Nucleic acids research. Jan. 8, 2019;47(D1):D766-73. |
Galon et al., Type, density, and location of immune cells within human colorectal tumors predict clinical outcome. Science. Sep. 29, 2006;313(5795):1960-4. |
George et al., Hemophilia B gene therapy with a high-specific-activity factor IX variant. New England Journal of Medicine. Dec. 7, 2017;377(23):2215-27. |
Griffiths et al., Detection and removal of barcode swapping in single-cell RNA-seq data. Nature communications. Jul. 10, 2018;9(1):1-6. |
Hao et al., Fast and robust deconvolution of tumor infiltrating lymphocyte from expression profiles using least trimmed squares. PLoS computational biology. May 6, 2019;15(5):e1006976. 21 pages. |
Hirata et al., Tumor microenvironment and differential responses to therapy. Cold Spring Harbor perspectives in medicine. Jul. 1, 2017;7(7):a026781. 14 pages. |
Hoek et al., A cell-based systems biology assessment of human blood to monitor immune responses after influenza vaccination. PloS one. Feb. 23, 2015;10(2):e0118528. 24 pages. |
Holik et al., RNA-seq mixology: designing realistic control experiments to compare protocols and analysis methods. Nucleic acids research. Mar. 17, 2017;45(5):e30. 18 pages. |
International Search Report and Written Opinion for International Application No. PCT/US2021/022155 dated Jul. 5, 2021. |
Izar et al., A Single-Cell Landscape of High-Grade Serous Ovarian Cancer. Nature medicine. 2020;26:1271-1279. 23 pages. |
Ke et al., Lightgbm: A highly efficient gradient boosting decision tree. Advances in neural information processing systems (NIPS). 2017;30:3146-54. |
Lambrechts et al., Phenotype molding of stromal cells in the lung tumor microenvironment. Nature medicine. Aug. 2018;24(8):1277-89. 19 pages. |
Levine et al., Data-driven phenotypic dissection of AML reveals progenitor-like cells that correlate with prognosis. Cell. Jul. 2, 2015;162(1):184-97. 31 pages. |
Linsley et al., Copy number loss of the interferon gene cluster in melanomas is linked to reduced T cell infiltrate and poor patient prognosis. PloS one. Oct. 14, 2014;9(10):e109760. 9 pages. |
Lun et al., EmptyDrops: distinguishing cells from empty droplets in droplet-based single-cell RNA sequencing data. Genome biology. Dec. 2019;20(1):1-9. |
Ma et al., PD1 Hi CD8+ T cells correlate with exhausted signature and poor clinical outcome in hepatocellular carcinoma. Journal for immunotherapy of cancer. Dec. 2019;7(1):331. 15 pages. |
Macosko et al., Highly parallel genome-wide expression profiling of individual cells using nanoliter droplets. Cell. May 21, 2015;161(5):1202-14. 25 pages. |
Marioni et al., RNA-seq: an assessment of technical reproducibility and comparison with gene expression arrays. Genome research. Sep. 1, 2008;18(9):1509-17. |
Melsted et al., Modular and efficient pre-processing of single-cell RNA-seq. BioRxiv. Jun. 17, 2019:673285. 16 pages. |
Monaco et al., RNA-Seq signatures normalized by mRNA abundance allow absolute deconvolution of human immune cell types. Cell reports. Feb. 5, 2019;26(6):1627-40.e7. |
Neftel et al., An integrative model of cellular states, plasticity, and genetics for glioblastoma. Cell. Aug. 8, 2019;178(4):835-49.e29. 37 pages. |
Newman et al. Robust enumeration of cell subsets from tissue expression profiles Nature Methods vol. 12, pp. 453-457 (Year: 2015). * |
Newman et al., CIBERSORT. Stanford University. 2021. https://cibersort.stanford.edu/ [last accessed Jun. 7, 2021]. 1 page. |
Newman et al., CIBERSORTx. Stanford University. 2021. https://cibersortx.stanford.edu/ [last accessed Jun. 7, 2021]. 1 page. |
Newman et al., Determining cell type abundance and expression from bulk tissues with digital cytometry. Nature biotechnology. Jul. 2019;37(7):773-82. |
Newman et al., Robust enumeration of cell subsets from tissue expression profiles. Nature methods. May 2015;12(5):453-7. 20 pages. |
Norton et al., Pancreatic cancer associated fibroblasts (CAF): under-explored target for pancreatic cancer treatment. Cancers. May 2020;12(5):1347. 18 pages. |
Puram et al., Single-cell transcriptomic analysis of primary and metastatic tumor ecosystems in head and neck cancer. Cell. Dec. 14, 2017;171(7):1611-24.e24. 40 pages. |
Racle et al., EPIC: a tool to estimate the proportions of different cell types from bulk gene expression data. Bioinformatics for Cancer Immunotherapy. 2020:233-248. |
Racle et al., Simultaneous enumeration of cancer and immune cell types from bulk tumor gene expression data. elife. Nov. 13, 2017;6:e26476. 25 pages. |
Rakaee et al.,Prognostic value of macrophage phenotypes in resectable non-small cell lung cancer assessed by multiplex immunohistochemistry. Neoplasia. Mar. 1, 2019;21(3):282-93. |
Roider et al., Dissecting intratumour heterogeneity of nodal B-cell lymphomas at the transcriptional, genetic and drug-response levels. Nature Cell Biology. Jul. 2020;22(7):896-906. 27 pages. |
Saltz et al., Spatial organization and molecular correlation of tumor-infiltrating lymphocytes using deep learning on pathology images. Cell reports. Apr. 3, 2018;23(1):181-93.e7. 21 pages. |
Shin et al., Variation in RNA-Seq transcriptome profiles of peripheral whole blood from healthy individuals with and without globin depletion. PloS one. Mar. 7, 2014;9(3):e91041. 11 pages. |
Stuart et al., Comprehensive integration of single-cell data. Cell. Jun. 13, 2019;177(7):1888-902.e21. 52 pages. |
Sturm et al., Comprehensive evaluation of transcriptome-based cell-type quantification methods for immuno-oncology. Bioinformatics. Jul. 15, 2019;35(14):i436-45. |
Sun et al. An Efficient and Flexible Method for Deconvoluting Bulk RNA-Seq Data with Single-Cell RNA-Seq Data Cells vol. 8 article 1161 (Year: 2019). * |
Tirosh et al., Dissecting the multicellular ecosystem of metastatic melanoma by single-cell RNA-seq. Science. Apr. 8, 2016;352(6282):189-96. |
Van Gassen et al., FlowSOM: Using self-organizing maps for visualization and interpretation of cytometry data. Cytometry Part A. Jul. 2015;87(7):636-45. |
Vivian et al., Toil enables reproducible, open source, big biomedical data analyses. Nature biotechnology. Apr. 2017;35(4):314-6. |
Wagner et al., Measurement of mRNA abundance using RNA-seq data: RPKM measure is inconsistent among samples. Theory in biosciences. Dec. 1, 2012;131(4):281-5. |
Wikipedia Nonlinear Regression retrieved from the internet on May 20, 2021 https://en.wikipedia.org/wiki/Nonlinear_regression (Year : 2021). * |
Wu et al., Stromal PD-L1-positive regulatory T cells and PD-1-positive CD8-positive T cells define the response of different subsets of non-small cell lung cancer to PD-1/PD-L1 blockade immunotherapy. Journal of Thoracic Oncology. Apr. 1, 2018;13(4):521-32. |
Xu et al., Mapping of γ/δ T cells reveals Vδ2+ T cells resistance to senescence. EBioMedicine. Jan. 1, 2019;39:44-58. |
Zaitsev et al., Complete deconvolution of cellular mixtures based on linearity of transcriptional signatures. Nature communications. May 17, 2019;10(1):2209. 16 pages. |
Zheng et al., Massively parallel digital transcriptional profiling of single cells. Nature communications. Jan. 16, 2017;8(1):1-12. |
Zimmermann et al., System-wide associations between DNA-methylation, gene expression, and humoral immune response to influenza vaccination. PloS one. Mar. 31, 2016;11(3):e0152034. 21 pages. |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20220230707A1 (en) * | 2020-03-12 | 2022-07-21 | Bostongene Corporation | Systems and methods for deconvolution of expression data |
US11587642B2 (en) * | 2020-03-12 | 2023-02-21 | Bostongene Corporation | Systems and methods for deconvolution of expression data |
Also Published As
Publication number | Publication date |
---|---|
IL296316A (en) | 2022-11-01 |
WO2021183917A1 (en) | 2021-09-16 |
AU2021233926A1 (en) | 2022-09-29 |
US20220230707A1 (en) | 2022-07-21 |
EP4118657B1 (en) | 2024-05-01 |
EP4118657A1 (en) | 2023-01-18 |
US20230178178A1 (en) | 2023-06-08 |
EP4383262A2 (en) | 2024-06-12 |
US20210287759A1 (en) | 2021-09-16 |
US11587642B2 (en) | 2023-02-21 |
JP2023518185A (ja) | 2023-04-28 |
CA3175126A1 (en) | 2021-09-16 |
WO2021183917A8 (en) | 2021-12-16 |
WO2021183917A9 (en) | 2021-11-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11315658B2 (en) | Systems and methods for deconvolution of expression data | |
US11802314B2 (en) | Methods and systems for determining proportions of distinct cell subsets | |
Ravi et al. | T-cell dysfunction in the glioblastoma microenvironment is mediated by myeloid cells releasing interleukin-10 | |
Mo et al. | Stromal gene expression is predictive for metastatic primary prostate cancer | |
Guiu et al. | Molecular subclasses of breast cancer: how do we define them? The IMPAKT 2012 Working Group Statement | |
Stavropoulou et al. | MLL-AF9 expression in hematopoietic stem cells drives a highly invasive AML expressing EMT-related genes linked to poor outcome | |
US20200395097A1 (en) | Pan-cancer model to predict the pd-l1 status of a cancer cell sample using rna expression data and other patient data | |
JP7299169B2 (ja) | 体細胞突然変異のクローン性を決定するための方法及びシステム | |
CN112602156A (zh) | 用于检测残留疾病的系统和方法 | |
CN105917008A (zh) | 用于前列腺癌复发的预后的基因表达面板 | |
CN109642259A (zh) | 使用肿瘤教育的血小板的针对癌症的群智能增强的诊断和治疗选择 | |
CN106755415A (zh) | 用于诊断和预测移植物排斥的生物标志物板 | |
JP2022538499A (ja) | サンプル調製、サンプルシークエンシング、およびシークエンシングデータのバイアス補正と品質管理のためのシステムならびに方法 | |
Barefoot et al. | Detection of cell types contributing to cancer from circulating, cell-free methylated DNA | |
US20230073731A1 (en) | Gene expression analysis techniques using gene ranking and statistical models for identifying biological sample characteristics | |
CN106661634A (zh) | 用于诊断肾异体移植物纤维化和排异风险的方法 | |
Jin et al. | Cross-species gene expression analysis reveals gene modules implicated in human osteosarcoma | |
CN101400804B (zh) | 用于结肠直肠癌预后的基因表达标志物 | |
Kastenschmidt et al. | A human lymphoma organoid model for evaluating and targeting the follicular lymphoma tumor immune microenvironment | |
Tuggle et al. | Methods for transcriptomic analyses of the porcine host immune response: application to Salmonella infection using microarrays | |
Lu et al. | Integrated identification of disease specific pathways using multi-omics data | |
Neary et al. | Methylated DNA immunoprecipitation sequencing (MeDIP-seq): Principles and applications | |
Zhang et al. | Comprehensive analysis of a novel RNA modifications-related model in the prognostic characterization, immune landscape and drug therapy of bladder cancer | |
US20220223227A1 (en) | Machine learning techniques for identifying malignant b- and t-cell populations | |
Lambrechts et al. | Differences in the tumor molecular and microenvironmental landscape between early (non-metastatic) and de novo metastatic primary luminal breast tumors |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
FEPP | Fee payment procedure |
Free format text: ENTITY STATUS SET TO UNDISCOUNTED (ORIGINAL EVENT CODE: BIG.); ENTITY STATUS OF PATENT OWNER: SMALL ENTITY |
|
FEPP | Fee payment procedure |
Free format text: ENTITY STATUS SET TO SMALL (ORIGINAL EVENT CODE: SMAL); ENTITY STATUS OF PATENT OWNER: SMALL ENTITY |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: NON FINAL ACTION MAILED |
|
AS | Assignment |
Owner name: BOSTONGENE CORPORATION, MASSACHUSETTS Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:BOSTONGENE LLC;REEL/FRAME:058273/0181 Effective date: 20211117 Owner name: BOSTONGENE LLC, RUSSIAN FEDERATION Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:ZAITSEV, ALEKSANDR;CHELUSHKIN, MAKSIM;CHEREMUSHKIN, ILYA;AND OTHERS;REEL/FRAME:058273/0110 Effective date: 20211111 |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: RESPONSE TO NON-FINAL OFFICE ACTION ENTERED AND FORWARDED TO EXAMINER |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: NOTICE OF ALLOWANCE MAILED -- APPLICATION RECEIVED IN OFFICE OF PUBLICATIONS |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: PUBLICATIONS -- ISSUE FEE PAYMENT VERIFIED |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: PUBLICATIONS -- ISSUE FEE PAYMENT VERIFIED |
|
STCF | Information on status: patent grant |
Free format text: PATENTED CASE |